├── README.md
├── data
    ├── small-dev.txt
    ├── small-dev.unk.txt
    ├── small-test.txt
    ├── small-test.unk.txt
    ├── small-train.txt
    ├── small-train.unk.txt
    └── vocab.txt
├── tutorial_bilstm_tagger.py
├── tutorial_bilstm_viterbi_tagger.py
├── tutorial_parser.ipynb
├── tutorial_rnnlm.py
├── tutorial_rnnlm_minibatch.py
└── tutorial_transition_parser.py


/README.md:
--------------------------------------------------------------------------------
 1 | # Practical Neural Networks for NLP
 2 | 
 3 | A tutorial given by [Chris Dyer](http://www.cs.cmu.edu/~cdyer/), [Yoav Goldberg](https://www.cs.bgu.ac.il/~yoavg/uni/), and [Graham Neubig](http://www.phontron.com/) at EMNLP 2016 in Austin. The tutorial covers the basic of neural networks for NLP, and how to implement a variety of networks simply and efficiently in the [DyNet](https://www.github.com/clab/dynet) toolkit.
 4 | 
 5 | * [Slides, part 1: Basics](http://phontron.com/slides/emnlp2016-dynet-tutorial-part1.pdf)
 6 |     * Computation graphs and their construction
 7 |     * Neural networks in DyNet
 8 |     * Recurrent neural networks
 9 |     * Minibatching
10 |     * Adding new differentiable functions
11 | 
12 | * [Slides, part 2: Case studies in NLP](http://phontron.com/slides/emnlp2016-dynet-tutorial-part2.pdf)
13 |     * Tagging with bidirectional RNNs and character-based embeddings
14 |     * Transition-based dependency parsing
15 |     * Structured prediction meets deep learning
16 | 


--------------------------------------------------------------------------------
/data/small-test.unk.txt:
--------------------------------------------------------------------------------
  1 | Trading was heavy at about one billion shares , compared with ###.# million Friday . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
  2 | But the session was <unk> , in contrast to the market 's four-day <unk> after the #### crash . ||| SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
  3 | Richard <unk> , a director at Hong <unk> <unk> International Fund Managers Ltd. , said the market probably has n't hit bottom yet but is close . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
  4 | In Australia , <unk> 's All <unk> index closed at ####.# , down #.# % , its biggest drop since October #### . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
  5 | London 's Financial <unk> Exchange ###-share index , the most closely watched market barometer , ended at its <unk> high of ####.# , down ##.# , or #.# % . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
  6 | At its low , shortly before Wall Street opened , it was off more than ### points . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
  7 | The Financial Times ##-share index closed ##.# points lower at ####.# . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R
  8 | Volume more than doubled to ###.# million shares from ###.# million Friday . ||| SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R
  9 | Prices on the Frankfurt Stock Exchange tumbled in heavy trading . ||| SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
 10 | The decline in the German Stock Index of ###.## points , or ##.# % , to ####.## was the Frankfurt market 's <unk> fall ever . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R
 11 | <unk> investors dumped holdings on a massive scale , pushing some blue-chip shares down as much as ## % . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 12 | To make them directly comparable , each index is based on the close of #### equaling ### . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_L SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 13 | The percentage change is since year-end . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R
 14 | We 've always thought that Mr. Wright <unk> California 's <unk> , but maybe the state 's <unk> <unk> are starting to <unk> the forces that made it such a significant place . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 15 | <unk> , the Environmental Defense Fund is having nothing to do with this one . ||| SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 16 | Not only Californians but all Americans would pay if this thing passed . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R
 17 | The initiative <unk> the sale of any crops in California that do n't meet the initiative 's standards . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 18 | Kansas <unk> farmers and Florida fruit growers would have to adjust or give up the California market . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 19 | In other words , California is <unk> to take control of the nation 's farm policy . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 20 | Consider the <unk> provision . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R
 21 | Even if one buys into the whole greenhouse theory , it is <unk> that reductions in a single state could have any impact on what is <unk> as a global problem . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 22 | But if <unk> science and economics have nothing to do with the new environment initiative , what is going on ? ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R
 23 | The key here is the <unk> of state Attorney General John Van de <unk> . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
 24 | The initiative seems to have been <unk> to include all the hot issues that set off the wealthy Hollywood <unk> who donate money . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 25 | And it allows Mr. Van de <unk> to get around campaign spending limits . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 26 | This initiative is being <unk> The Big Green , but maybe it should be called The Big <unk> . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
 27 | -LRB- The Republican candidate , Sen. <unk> Wilson , is playing the initiative <unk> game too , <unk> his own crime initiative . -RRB- ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R
 28 | While it is possible that the Big Green initiative will be ruled <unk> , it is of course <unk> that in modern California it could slide through . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R
 29 | This is the state that recently passed the <unk> ## <unk> initiative . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
 30 | If this new proposal ever does become law , the green lobby will benefit directly . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R
 31 | The initiative <unk> a free floating state environmental officer to <unk> companies or government agencies that do things he does n't like . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 32 | That means the <unk> and such groups no longer would have to spend as much money on litigation ; taxpayers would bear the cost . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
 33 | Of course the state 's <unk> are not yet a nation <unk> themselves . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R
 34 | Then <unk> . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R
 35 | <unk> by <unk> of large-scale buying in blue-chip stocks , the Dow Jones Industrial Average rallied yesterday and <unk> about a half of Friday 's <unk> plunge , gaining ##.## to ####.## . ||| SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 36 | While the advance cheered investors who feared a <unk> crash would occur yesterday , it was strictly a <unk> rally fed by huge buying by <unk> institutions and program traders . ||| SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 37 | The Nasdaq OTC index closed down #.## to ###.## . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 38 | <unk> plunged on takeover <unk> in two airline stocks , UAL and AMR , which each fell more than ## % when they <unk> for trading yesterday after being suspended Friday afternoon . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 39 | Overall , `` this is a <unk> rally but it 's very <unk> , '' said Arthur <unk> Jr. , a veteran PaineWebber Inc. trader at the Big Board . ||| SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 40 | It 's just a <unk> feeling . ||| SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
 41 | I do n't think anyone left the place <unk> <unk> . '' ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R
 42 | But to traders , it looked like disaster on the #:## a.m. opening bell . ||| SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
 43 | At #:## , Procter & Gamble -- one of the most important Dow <unk> of late -- opened down # #/# to ### . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 44 | The Dow dropped to a quick <unk> loss , and to many traders it looked as if stocks were headed for yet another big tumble . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 45 | Then , to make matters worse , computerized sell programs <unk> in , <unk> stocks into <unk> losses . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 46 | There was heavy stock-index arbitrage , as traders sold big <unk> of stock and bought stock-index futures to profit from the price <unk> between the two markets . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 47 | This was a <unk> from Friday , when Standard & Poor 's ###-stock index futures had closed at a sharp discount to stocks . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 48 | They did n't . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R
 49 | The Dow accelerated its slide , losing ##.## in the first ## minutes of trading . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 50 | Then at ##:## the Dow suddenly started to rebound , and when it shot upward it did so even faster than the <unk> fall . ||| SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 51 | All the selling had pushed stocks to such cheap values that big investment banks and major money management firms started buying stocks heavily . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 52 | Trading in Walt Disney Co. particularly caught traders ' eyes . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R
 53 | According to Big Board officials , Disney had one of the biggest <unk> <unk> on Friday ; it was one of the seven stocks that could n't finish trading that day . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 54 | But then it shot upward # #/# as Goldman , Sachs & Co. stepped in and bought , traders said . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
 55 | <unk> Wall Street , trading desks were <unk> that they could at least play the market yesterday , in contrast to Friday 's gridlock . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 56 | We put some orders together . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R
 57 | There was n't a lot of panic selling , either <unk> or <unk> ... . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R
 58 | Not like Friday where they just took -LCB- the market -RCB- apart . '' ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R
 59 | <unk> <unk> Peck , a trader in Shearson Lehman Hutton Inc. 's OTC department : `` I tell you , this market acts healthy . '' ||| SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R
 60 | <unk> him , <unk> of traders seemed to get a burst of energy ; their <unk> broke out bottles of <unk> water to cool them off . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 61 | Among Big Board specialists , the <unk> was `` <unk> your offers '' -- meaning that specialists soon expected to get higher prices for their shares . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 62 | But not everybody was making money . ||| SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R
 63 | The <unk> on the Chicago Board Options Exchange , the nation 's major options market , was heavy after the trading in S&P ### stock-index options was halted Friday . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R
 64 | `` They were <unk> killed , <unk> , '' said one Chicago-based options trader . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R
 65 | Some traders said that the closely watched Major Market Index , whose ## stocks mimic the Dow industrials , did n't lead yesterday 's big rally . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
 66 | Then when the market was at a technical level to buy , they came in with a vengeance . '' ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R
 67 | However , according to one analyst , the timing of Major Market Index futures buying just before the turnaround was similar to that of <unk> Tuesday . ||| SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 68 | `` Futures were pulling the stock market higher , '' said Donald <unk> , head of stock-index futures research at Prudential-Bache Securities Inc . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 69 | <unk> were criticized for their inability to maintain <unk> markets during the Friday plunge . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 70 | According to a Big Board official , while many stocks opened late , there were subsequent trading <unk> in only three issues -- AMR , Merck and <unk> Energy . ||| SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 71 | Merck is one of the most important stocks in the Major Market Index . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 72 | No sector of the market has been <unk> during the past two days ' gyrations . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
 73 | For example , <unk> jumped # #/# to ## , Schlumberger rose # #/# to ## #/# and Baker Hughes rose # #/# to ## . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 74 | Because of the UAL and AMR <unk> , airlines were the weakest sector of the market yesterday . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R
 75 | Philip Morris was the Big Board 's most active issue , rising # #/# to ## #/# on nearly eight million shares . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 76 | Merrill Lynch added # #/# to ## , PaineWebber rose #/# to ## #/# and Bear Stearns rose #/# to ## #/# . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 77 | Federal National Mortgage Association , a recently hot stock , climbed # to ### on nearly #.# million shares . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
 78 | Mr. Phelan said that program trading strategies were n't responsible for <unk> Friday 's decline despite a jump in the use of the computer-driven strategies in recent months . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 79 | Mr. Phelan expressed relief that the market rebounded yesterday . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 80 | `` It was the opposite of what happened on Oct. ## . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 81 | They used their judgment . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R
 82 | Instead , they bought on weakness and sold into the strength , which kept the market <unk> . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
 83 | Maybe they learned from experience . '' ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R
 84 | Wall Street traders on Friday had complained about the trading <unk> . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
 85 | James A. White and <unk> <unk> contributed to this article . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
 86 | West Germany 's Green Party joined its ideological <unk> <unk> <unk> and the <unk> Institute in the legal battle to ground the Atlantis shuttle and its <unk> Galileo <unk> to <unk> . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 87 | Of course it was . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R
 88 | NASA should now <unk> for fines against all three <unk> , foreign and domestic , for bringing this <unk> case . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 89 | A House-Senate conference approved a permanent smoking ban on all domestic airline routes within the <unk> U.S. and on all flights of six hours or less to Alaska and <unk> . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 90 | The restrictions would cover all but a small percentage of domestic air traffic , and represent a major expansion of the current smoking ban on flights of two hours or less . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 91 | From the <unk> , the tobacco industry has been uncertain as to what strategy to follow . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 92 | `` <unk> can vote as they want , '' said Rep. William Lehman -LRB- D. , <unk> -RRB- , head of the House conferees . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 93 | <unk> of the costs -- and , critics say , competition -- the airlines have sought to gain leverage over the city of Denver . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 94 | The industry sought to impose conditions that would have delayed funds for the project until Denver and the airlines had agreed to <unk> for ## % of the gates . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 95 | Congress previously cut six airports this year . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R
 96 | And <unk> grants for mass <unk> would be effectively frozen at $ #.### billion , or $ ## million more than last fiscal year . ||| SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 97 | Sony two weeks ago agreed to acquire Columbia for $ #.# billion , or $ ## a share . ||| SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 98 | Warner sued Sony and <unk> late last week ; Sony and <unk> have <unk> , charging Warner with attempting to <unk> in Sony 's acquisition of the two companies . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
 99 | <unk> 's net income in the latest quarter compared with a net loss of $ #.# million , or ## cents a share , in the year-earlier period . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
100 | Officials at <unk> , based in Pittsburgh , declined comment . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R
101 | Congress has been critical of the Bush administration for not sending enough aid to Poland , so it is getting ready to send its own version of a <unk> package . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
102 | Last month , the Senate voted to send a delegation of congressional staffers to Poland to <unk> its legislature , the <unk> , in democratic procedures . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
103 | Senator <unk> <unk> calls this effort `` the first gift of democracy . '' ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R
104 | Maybe after the staffers explain their work to the Poles , they 'd be willing to come back and do the same for the American people . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
105 | In a sharply weaker London market yesterday , <unk> shares were down ## pence at ## pence -LRB- ## cents -RRB- . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
106 | There were n't any extraordinary items . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R
107 | Sales for the total group rose ## % to ###.# million <unk> <unk> compared with ###.# million <unk> <unk> a year ago . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
108 | <unk> has decided against paying an interim dividend . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
109 | <unk> said the appointment of a new management team and the signing of a comprehensive labor agreement are expected to enhance the company 's long-term prospects . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
110 | The sudden `` flight to quality '' that triggered Friday 's <unk> <unk> rally was reversed yesterday in a `` flight from quality '' <unk> . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
111 | `` It was a pretty wild day . ||| SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
112 | Our markets were closely tied to the stock market , '' said <unk> <unk> , manager of trading at Smith Barney , Harris Upham & Co . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
113 | `` Friday 's flight to quality was no longer needed once the stock market found its <unk> , '' he said . ||| SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R
114 | That caused investors to <unk> stocks and buy high-quality Treasury bonds , which are <unk> than other types of securities . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
115 | <unk> to the selling pressure were <unk> by several investment firms advising clients to boost their stock holdings and reduce the size of their cash or bond portfolios . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
116 | The rate is considered an early signal of changes in Fed policy . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
117 | In fact , some economists contend that the latest easing started last week . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
118 | The Treasury 's benchmark ##-year bond , ended about # #/# points lower , or down about $ ##.## for each $ #,### face amount . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
119 | After Treasury bill rates plummeted as much as #.## percentage point on Friday , they gave back <unk> of that amount yesterday . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R
120 | The bond-equivalent yield on three-month Treasury bills , for example , was quoted late yesterday at #.## % , compared with #.## % Friday . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
121 | Investment-grade corporate bonds , mortgage-backed securities and municipal bonds also fell . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R
122 | But prices of junk bonds , which were battered Friday in near standstill trading , rebounded to post small gains after a volatile trading session . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
123 | Junk bonds opened as much as four points lower but <unk> a modest comeback as stock prices firmed . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
124 | Some traders said the high-yield market was helped by active institutional buying . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
125 | The benchmark ##-year bond , for example rose one point in early Japanese trading in reaction to a quick ###-point drop in the Tokyo stock market . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
126 | But as Japanese stocks rebounded , Treasurys retreated and ended just modestly higher . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
127 | Many U.S. trading operations , <unk> to keep a <unk> eye on Japanese trading as an indication of where U.S. trading would begin , were fully <unk> during the Tokyo trading session . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
128 | `` Most of the action was during the night session , '' said Michael <unk> , trading manager at Continental Bank . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
129 | <unk> <unk> , who often trades overnight for Capital <unk> Inc. , Beverly Hills , Calif. , said trading in Tokyo was `` very active '' but highly volatile . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
130 | In Tokyo , trading is halted during <unk> . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R
131 | The August trade deficit is expected to have <unk> to $ #.# billion from $ #.## billion in July . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
132 | A <unk> of that <unk> , said one New York trader , is `` not a favorable number ... . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R
133 | It could do damage to us . '' ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R
134 | Resolution Funding is a division of Resolution Trust Corp. , the new federal agency created to <unk> out the nation 's troubled thrifts . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
135 | `` There 's lots of supply , '' the New York trader said . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
136 | `` We have a couple or three tough weeks coming . '' ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R
137 | The benchmark ##-year Treasury bond was quoted late at a price of ### ##/## , compared with a closing price of ### ##/## Friday . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
138 | The yield on the benchmark issue rose to #.## % from #.## % . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
139 | Short-term interest rates fell yesterday at the government 's weekly Treasury bill auction . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
140 | The average discount rate was #.## % on new six-month bills , the lowest since the average of #.## % at the auction on July ## , #### . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
141 | Here are auction details : ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
142 | Rates are determined by the difference between the purchase price and face value . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
143 | Thus , higher bidding <unk> the investor 's return while lower bidding <unk> it . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R
144 | Corporate Issues ||| SHIFT SHIFT REDUCE_L
145 | Investment-grade corporate bonds ended one to # #/# point lower . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R
146 | Foreign bonds surged as the dollar weakened against most major currencies . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
147 | The yield was #.### % . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
148 | Mortgage securities gave up most of Friday 's gains as active issues ended ##/## to ##/## point lower . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
149 | On Friday , mortgage issues gained as much as # #/## . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
150 | Late yesterday <unk> Mae # % securities were yielding #.## % to a ##-year average life assumption , as the spread above the Treasury ##-year note narrowed #.## percentage point to #.## . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
151 | Municipals ||| SHIFT
152 | <unk> stocks and weaker Treasury prices drove municipal bonds #/# to #/# point lower in late dealings . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
153 | The session losses left municipal dollar bonds close to where they were before the <unk> drop in the Dow Jones Industrial Average Friday prompted a capital markets rally . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
154 | <unk> dominated municipal trading throughout the session . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
155 | New Jersey Turnpike Authority 's #.## % issue of #### was off #/# at ## #/# bid , yielding #.## % , up #.## percentage point from late Friday . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
156 | Florida Board of Education 's # #/# % issue of #### was #/# point weaker at ## #/# bid . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
157 | The # #/# % issue of <unk> Bridge and <unk> Authority of New York , due #### , was off #/# at ## #/# bid . ||| SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
158 | <unk> , a security services and auctions company , trades on London 's Stock Exchange . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
159 | <unk> is <unk> by Canadian Pacific Ltd. , a <unk> transportation , resources and industrial holding concern . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
160 | <unk> net fell to ###.# yen from ###.# yen because of expenses and capital adjustments . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
161 | <unk> <unk> sales , however , were lower . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
162 | <unk> Corp. of Park Ridge , N.J. , said it retained Merrill Lynch Capital Markets to sell its <unk> Equipment <unk> Corp. unit . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
163 | `` We are only going to sell at the right price . '' ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R
164 | <unk> Equipment had operating profit before depreciation of $ ## million on revenue of $ ### million in #### . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
165 | The closely held <unk> Corp. had annual revenue of close to $ # billion in #### , of which $ #.# billion was contributed by its <unk> <unk> A Car operations world-wide . ||| SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
166 | It supplies commercial and industrial equipment including <unk> , <unk> , <unk> and electrical equipment , <unk> , <unk> , <unk> and trucks . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
167 | In the year-earlier quarter , the company reported net income of $ ###,### , or ## cents a share . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
168 | In over-the-counter trading , <unk> fell ##.# cents to $ #.## . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
169 | In the first nine months of #### , net was $ ## million , or $ #.## a share . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
170 | Mr. Simmons said the third-quarter results reflect continued improvements in productivity and operating margins . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
171 | U.S. <unk> said it believes the sale , if completed , apparently would <unk> antitrust issues raised by the U.S. Justice Department about U.S. <unk> 's offer to buy International <unk> . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
172 | Both of the New York-based companies <unk> stock certificates and currency . ||| SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R
173 | It also said the tender offer would probably have to be extended further to complete financing arrangements . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
174 | The offer , made June # , has been extended several times . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R
175 | U.S. <unk> said that as of Oct. ## , ##.# million shares , or about ##.# % of the fully diluted shares outstanding , had been tendered . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R
176 | <unk> Group Inc. said it agreed to buy ## % of <unk> <unk> Ltd. , a closely held apparel maker , with the assumption of $ # million of <unk> debt . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
177 | That ## % is now held by <unk> Parker , <unk> 's president and chief executive officer , who will continue to manage <unk> 's operations under <unk> . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
178 | In #### , <unk> will have sales `` in excess of $ ## million '' and will show a profit , Mr. Parker said . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
179 | <unk> , which makes <unk> apparel sold mainly through mass <unk> like K mart and <unk> , said the <unk> acquisition will enhance its strategy to expand into department stores . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
180 | This fall , <unk> began manufacturing moderately priced clothes aimed at department stores under the <unk> <unk> <unk> , which <unk> recently acquired . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
181 | Those results included a $ #.# million charge related to the retirement of debt . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
182 | Enron said each unit will be priced in the $ <unk> range and will represent about ## % of the partnership equity . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
183 | Arthur M. Goldberg said he extended his <unk> tender offer of $ ## a share tender offer , or $ ###.# million , for Di <unk> Corp. to Nov. # . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
184 | <unk> Acquisition Corp. , the New Jersey investor 's acquisition vehicle , said that as of the close of business yesterday , ###,### shares had been tendered . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R
185 | <unk> the stake <unk> already held , <unk> holds a total of about ## % of Di <unk> 's shares on a fully diluted basis . ||| SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
186 | The offer , which also includes common and preferred stock purchase rights , was to expire last night at midnight . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
187 | The new <unk> date is the date on which <unk> 's financing commitments , which total about $ ### million , are to expire . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
188 | <unk> is a unit of <unk> Holding Corp. , a unit of Rose Partners <unk> . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
189 | In New York Stock Exchange composite trading yesterday , Di <unk> closed at $ ##.## a share , down $ #.## . ||| SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
190 | A. <unk> <unk> , B. <unk> <unk> , C. radio <unk> shows . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R
191 | If you <unk> <unk> <unk> , you 're right . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
192 | After years of <unk> into the <unk> , <unk> <unk> is coming back . ||| SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R
193 | <unk> magazine advertisements feature <unk> <unk> <unk> of Hollywood <unk> <unk> <unk> , <unk> and liquor . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
194 | <unk> <unk> accustomed to <unk> only in color report a rush to <unk> <unk> orders . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
195 | What 's happening in <unk> <unk> the popularity of black and white in fashion , home <unk> and <unk> . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
196 | On <unk> Avenue , <unk> have been advancing the <unk> look with clothing <unk> done entirely in black and white . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
197 | And classic <unk> movies are <unk> a comeback on <unk> tapes , spurred , in part , by the <unk> against <unk> of old films . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
198 | `` The <unk> is <unk> back to black and white , '' says Richard <unk> , the general manager of Eastman Kodak Co. 's professional <unk> division . ||| SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
199 | Until two years ago , sales of <unk> film had been declining steadily since the ####s . ||| SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
200 | <unk> at commercial <unk> , the film can be used in very low light without <unk> quality , says Donald <unk> of <unk> <unk> . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
201 | <unk> recently signed <unk> gold <unk> Florence <unk> to <unk> a new line of <unk> paper that 's <unk> to consumers and will compete directly with Kodak 's papers . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
202 | The biggest <unk> of the <unk> <unk> is likely to be International Paper Co. 's <unk> division , known in the industry for its premium products . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
203 | Sales of <unk> 's four <unk> of <unk> film this year are <unk> growth in the overall market , although the company wo n't say by exactly how much . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
204 | `` We hope the trend <unk> , '' says <unk> <unk> , <unk> 's marketing communications director . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
205 | `` It has an <unk> , almost <unk> quality to it , '' says Owen B. <unk> , the chairman of the <unk> <unk> department at <unk> Institute of Technology . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
206 | `` You can shift out of reality with black and white , '' he adds . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R
207 | Such features have been especially attractive to professional <unk> and marketing executives , who have been steadily increasing their use of black and white in advertising . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
208 | Consider <unk> Inc. , whose latest ad campaign features <unk> shots of Hollywood stars , artists and other well-known <unk> <unk> the retailer 's <unk> and <unk> . ||| SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
209 | Richard <unk> , the account manager for the campaign , says <unk> did n't intentionally choose black and white to <unk> its ads from the color spreads of competitors . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
210 | `` We wanted to <unk> the individual , not the environment , '' he says , `` and black and white allows you to do that better than color . '' ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R
211 | The campaign won a <unk> award as this year 's best ad by a specialty retailer . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
212 | Even food products and <unk> , which have long <unk> on color , are making the switch . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R
213 | Other companies that are currently using <unk> ads include American Express Co. and <unk> America Inc . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
214 | <unk> <unk> have also <unk> onto the trend . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
215 | His <unk> <unk> studio in <unk> , <unk> , doubled its business last year and , he says , is <unk> solid for the next five . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
216 | One customer , <unk> <unk> , says she <unk> a color <unk> for black and white because `` it 's more dramatic . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
217 | I show it to my friends , and they all say ` <unk> . ' ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R
218 | It is n't ordinary like color . '' ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R
219 | Still , most consumers are n't <unk> <unk> film into their cameras to take family <unk> . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
220 | <unk> , it must be <unk> to a handful of processors and may take a week or more to be processed and returned . ||| SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
221 | But for <unk> , developing costs for <unk> film are higher . ||| SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
222 | Some companies are starting to tackle that problem . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
223 | <unk> , for example , recently introduced a <unk> film that can be processed quickly by color <unk> . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
224 | <unk> on <unk> customers , the company is also increasing its <unk> of <unk> <unk> classes . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
225 | Similarly , <unk> is <unk> <unk> of <unk> <unk> at high schools and colleges , offering free <unk> film and paper as prizes . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
226 | And Kodak is <unk> an <unk> video to processors on how to develop its <unk> film more <unk> . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
227 | Other companies are introducing related products . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R
228 | Charles <unk> Co. , a leading maker of <unk> <unk> , introduced last month a complete <unk> <unk> <unk> targeted at <unk> who want to process their own <unk> photographs . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
229 | But some industry observers believe the <unk> of black and white is only a fad . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R
230 | They <unk> the <unk> of still electronic <unk> , more newspapers turning to color on their pages and <unk> improvements in the quality of color <unk> . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
231 | `` Black and white has n't made the same <unk> <unk> in <unk> development as color , '' says Mr. <unk> of the <unk> Institute . ||| SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
232 | `` The color <unk> today is far superior to <unk> of ## years ago . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
233 | `` It 's got a classic spirit and carries over <unk> , '' says <unk> <unk> of Professional <unk> of America . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
234 | `` That 's the appeal . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
235 | Sales rose more than # % to $ ##.# million from $ ##.# million . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
236 | The Sacramento , Calif. , company also attributed improved performance to a lower effective tax rate and higher interest income . ||| SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
237 | Sales grew almost # % to $ ###.# million from $ ###.# million . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
238 | <unk> publishes the Sacramento -LRB- Calif . -RRB- <unk> and <unk> -LRB- Wash . -RRB- News Tribune , and other papers in Western states . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
239 | <unk> <unk> . A. and Societe National <unk> <unk> , the state oil companies of Italy and France , respectively , <unk> an offer to buy <unk> Suisse S.A . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
240 | The price was n't disclosed . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
241 | A spokesman for <unk> said that the Swiss oil concern was examining the offer , <unk> last Friday , along with two other offers , also <unk> last week . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
242 | Those two offers were private and the spokesman refused to identify the bidding companies . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
243 | The spokesman further said that at least two more offers are expected from other companies within two weeks . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
244 | The latest reading of ###.# was up from ###.# in July and ###.# as recently as March . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
245 | The August rise marked the fifth straight monthly gain for the indicator , which uses the #### average as a base of ### . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
246 | In contrast , the Commerce Department 's widely followed index of leading indicators , while up in August , has fallen repeatedly since reaching a high early this year . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
247 | But the far stronger showing of the Columbia index `` makes a recession any time soon highly unlikely , '' says <unk> H. <unk> , the director of the Columbia facility . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
248 | The group normally <unk> only when a change in the economy 's general course seems likely . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R
249 | `` No meeting is scheduled because the expansion shows no sign of going off the tracks , '' Mr. <unk> reports . ||| SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R
250 | The comparable lead times for the Commerce index , whose components include the stock market , are far <unk> -- ## months before <unk> and only three months before <unk> . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
251 | `` It was an entirely different pattern from what we 're seeing now , '' Mr. <unk> says . ||| SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R
252 | The stock market has lost some <unk> power , analysts at the Columbia center claim , because of the growing impact of international developments . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
253 | <unk> S.A. , a leading French food group , said it agreed to acquire <unk> <unk> . H. , a West German <unk> maker . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
254 | The value of the acquisition was n't disclosed . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
255 | The move is in line with <unk> 's strategy of gradually building its share of the European <unk> market through <unk> growth . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
256 | <unk> will initially acquire a ## % interest in <unk> , a closely held concern . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
257 | The French group has an agreement giving it the right to buy all the shares outstanding , and this could be completed within a few months , a <unk> spokeswoman said . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R
258 | The acquisition <unk> <unk> 's position in the European <unk> market . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
259 | The French group currently ranks second after <unk> Group of Italy , whose sales are <unk> in the Italian market . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
260 | The agency said it confirmed American Continental 's preferred stock rating at C . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
261 | <unk> <unk> BANK -LRB- Miami -RRB- -- ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R
262 | The rating agency also reduced the ratings for long-term deposits to <unk> from <unk> and for preferred stock to <unk> from <unk> . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
263 | THE <unk> <unk> <unk> a repeat of Black Monday as prices recovered from an early slide , spurred by <unk> institutions and program traders . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
264 | The rally <unk> about half of Friday 's <unk> plunge , but analysts are cautious about the market 's outlook . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
265 | The dollar also rebounded , while bond prices plummeted and Treasury bill rates soared . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
266 | Junk bonds also recovered somewhat , though trading remained stalled . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R
267 | Gold also rose . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R
268 | Tokyo stock prices bounced back in early trading Tuesday following a #.# % plunge on Monday . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
269 | The dollar also moved higher in Tokyo . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R
270 | AMR slid $ ##.### , to $ ##.## . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
271 | Also , a UAL group tried to get financing for a lower bid , possibly $ ### a share . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
272 | UAL fell $ ##.### , to $ ###.### . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
273 | IBM 's earnings tumbled ## % in the third quarter , slightly more than expected . ||| SHIFT SHIFT REDUCE_R SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
274 | The computer giant partly cited a stronger dollar and a delay in shipping a new <unk> disk drive . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
275 | <unk> at <unk> and managed plants in the U.S. is due to rise ## % . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
276 | Budget director Darman said he wo n't give federal agencies much <unk> in <unk> with Gramm-Rudman spending cuts , which took effect yesterday . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
277 | Darman hopes to <unk> Congress to finish a deficit plan . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
278 | The <unk> and <unk> plan would create another possible <unk> to selling sick thrifts . ||| SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
279 | The Supreme Court agreed to decide whether a federal court may <unk> a merger that has won regulatory approval but been ruled <unk> in a private suit . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
280 | Merrill Lynch 's profit slid ## % in the third quarter . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
281 | Bear Stearns posted a #.# % gain , while PaineWebber had a decline due to a year-ago gain . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
282 | <unk> <unk> of Britain plans to return to the name <unk> and take a big <unk> . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
283 | <unk> 's profit more than doubled . ||| SHIFT SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R
284 | K mart agreed to acquire <unk> <unk> <unk> for $ ### million , expanding its presence in the growing <unk> business . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
285 | Markets -- ||| SHIFT SHIFT REDUCE_R
286 | Dow Jones industrials ####.## , up ##.## ; transportation ####.## , off ###.## ; utilities ###.## , up #.## . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
287 | Commodities : Dow Jones futures index ###.## , off #.## ; spot index ###.## , up #.## . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
288 | Dollar : ###.## yen , off #.## ; #.#### marks , off #.#### . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
289 | Monday , October ## , #### ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R
290 | The key U.S. and foreign annual interest rates below are a guide to general levels but do n't always represent actual transactions . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
291 | The base rate on corporate loans at large U.S. money center commercial banks . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
292 | Reserves traded among commercial banks for overnight use in amounts of $ # million or more . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
293 | Source : Fulton Prebon -LRB- U.S.A . -RRB- Inc . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R
294 | The charge on loans to depository institutions by the New York Federal Reserve Bank . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
295 | CALL MONEY : # #/# % to ## % . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R
296 | The charge on loans to brokers on stock exchange collateral . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
297 | COMMERCIAL PAPER : High-grade unsecured notes sold through dealers by major corporations in multiples of $ #,### : #.## % ## days ; #.## % ## days ; #.## % ## days . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
298 | Average of top rates paid by major New York banks on primary new issues of negotiable C.D.s , usually on amounts of $ # million and more . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
299 | <unk> rates in the secondary market : #.## % one month ; #.## % three months ; #.## % six months . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
300 | BANKERS ACCEPTANCES : #.## % ## days ; #.## % ## days ; #.## % ## days ; #.## % ### days ; #.## % ### days ; #.## % ### days . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
301 | Negotiable , bank-backed business credit instruments typically financing an import order . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
302 | The average of interbank offered rates for dollar deposits in the London market based on quotations at five major banks . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
303 | FOREIGN PRIME RATES : Canada ##.## % ; Germany #.## % ; Japan #.### % ; Switzerland #.## % ; Britain ## % . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
304 | These rate indications are n't directly comparable ; lending practices vary widely by location . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
305 | FEDERAL HOME LOAN MORTGAGE CORP . -LRB- Freddie Mac -RRB- : Posted yields on ##-year mortgage commitments for delivery within ## <unk> ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R
306 | #.## % , standard conventional <unk> mortgages ; #.### % , # % rate capped one-year adjustable rate mortgages . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
307 | MERRILL LYNCH READY ASSETS TRUST : #.## % . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R
308 | <unk> average rate of return after expenses for the past ## days ; not a forecast of future returns . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
309 | Under the agreement , Intel will invest $ # million to acquire a # % stake in <unk> , a maker of <unk> for scientists and engineers . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
310 | <unk> said it plans to use the <unk> in future products . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
311 | In #### the Cincinnati company earned $ #.# million , or ## cents a share , on revenue of $ ###.# million . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
312 | <unk> said the division operates under the trade name <unk> and supplies the medical and <unk> markets . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
313 | Burmah Oil PLC , a British independent oil and specialty chemicals marketing concern , said <unk> Holdings N.V. of the Netherlands has built up a #.# % stake in the company . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
314 | James Alexander , a Burmah spokesman , said <unk> had previously owned `` a little under # % '' of Burmah for about two years . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
315 | The Dutch company had n't notified Burmah of its reason for increasing the stake , he said . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
316 | Burmah , which owns the <unk> brand of <unk> oils , reported a ## % rise in net income to # ##.# million -LRB- $ ##.# million -RRB- in the first half . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
317 | The group consists of <unk> Corp. and <unk> <unk> Inc . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
318 | The company 's remaining business is the manufacture and sale of engine and <unk> products for industrial and transportation applications . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
319 | <unk> a $ #.# million provision for <unk> accounts , <unk> National <unk> Inc. posted a loss for its fourth quarter ended June ## . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
320 | The company said the $ #.# million reserve was created to reflect doubt about the <unk> of <unk> owed to National <unk> by some of the real estate partnerships it manages . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
321 | The company also said expenses incurred by the previous board and management in the recent contest for control were recognized primarily in the first quarter ended Sept. ## . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
322 | `` The new structure will enable United <unk> to focus clearly upon opportunities for planned growth during the ####s , '' said Bob <unk> , deputy chairman and group chief executive . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
323 | Last month , United <unk> agreed to sell its entire restaurant operations to Grand Metropolitan PLC for # ### million . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
324 | An American <unk> now is standing trial in <unk> . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R
325 | The most likely winner will be the <unk> <unk> rebels . ||| SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
326 | The U.S. <unk> 's `` crime '' was writing that the head of the commission charged with <unk> the election 's fairness , <unk> <unk> , was openly <unk> to <unk> . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
327 | <unk> after that , Mr. <unk> had Scott Stanley <unk> and his <unk> confiscated . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
328 | <unk> has <unk> favorable Western media treatment ever since the U.N. General Assembly declared it the `` sole , <unk> representative '' of <unk> 's people in ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R
329 | The elections are set for Nov. # . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
330 | In July , Mr. Stanley , editor of American Press International , a Washington <unk> conservative wire service , visited <unk> to report on the <unk> election campaign . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
331 | After Mr. Stanley 's article was published in two <unk> newspapers , Mr. <unk> had criminal charges brought against their <unk> , publisher and lawyer . ||| SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
332 | Mr. Stanley was <unk> and charged along with the others when he returned to <unk> this month . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
333 | Both the State Department and the Lawyers Committee for Freedom of the Press have <unk> Mr. Stanley 's <unk> . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R
334 | Both South African and <unk> <unk> are intimidating voters . ||| SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R
335 | It now has the chance to <unk> that record in <unk> . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
336 | Commodity futures prices generally reflected the stability of the stock market following its plunge Friday . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
337 | Yesterday , the stock market 's influence at first created <unk> . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R
338 | Trading in cotton and sugar was nervous and showed small declines . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
339 | In Chicago , grain and <unk> prices rose slightly . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R
340 | <unk> and meat prices , however , dropped on concern that a financial crisis would cut consumption of beef and pork . ||| SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
341 | The spot October gold price rose $ # to $ ###.## an ounce . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
342 | December silver was up #.# cents an ounce at $ #.### . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
343 | <unk> <unk> more like an industrial metal , easing early on concern over a possible weaker economy , but <unk> later , as the stock market strengthened . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
344 | For one thing , last Friday , precious metals markets closed before the stock market went into its <unk> nose <unk> , so it could n't <unk> to it . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
345 | Back on Friday , Oct. ## , <unk> , the stock market declined during the day , and gold prices surged as the stock market fell . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
346 | Yesterday 's October gain of $ # was <unk> compared with that . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
347 | There 's a good chance that gold will retain its gains and rise further , he said . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
348 | He expects a drop in interest rates , which would help gold by keeping the dollar from rising . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
349 | Finally , according to Mr. <unk> , the impact of the strong dollar should be reflected in reduced exports in the August merchandise trade deficit , when the figures are released today . ||| SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R
350 | This would be <unk> to the dollar and <unk> for gold , he said . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
351 | <unk> : ||| SHIFT SHIFT REDUCE_R
352 | The U.S. benchmark crude , West Texas Intermediate , closed at $ ##.## a barrel for November delivery , down ## cents . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
353 | But most market observers agreed that Friday 's stock market drop is what <unk> spirits in the petroleum pits yesterday . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
354 | Until yesterday , futures prices had been headed up on expectations that world oil demand will continue to be strong . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
355 | The Organization of Petroleum <unk> <unk> increased its production ceiling for the fourth quarter based on projections of robust demand . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
356 | Indeed , after <unk> early in the trading day to Friday 's <unk> , futures prices firmed up again as traders took note of the stock market 's partial recovery yesterday . ||| SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R
357 | <unk> : ||| SHIFT SHIFT REDUCE_R
358 | Futures prices fell and showed little rebound as one major labor problem that had been <unk> prices appeared to be solved . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
359 | Prices were down from the <unk> of trading on concern that a drop in the stock market might create a weakened economy and a <unk> reduction in copper use . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
360 | <unk> Valley is a large Canadian producer and principal supplier to Japan , which recently began seeking copper elsewhere as its inventories <unk> . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
361 | Last week it was reported that company and union negotiations had <unk> the major hurdle , the <unk> out of work by the company . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
362 | Now , the analyst said , only minor points remain to be <unk> up . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R
363 | `` For all <unk> and purposes , an agreement appears to have been achieved , '' he said . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
364 | <unk> inventories in New York 's Commodity Exchange warehouses rose yesterday by ### tons , to ##,### tons . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
365 | London Metal Exchange copper inventories last week declined ##,### tons , to ##,### tons . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
366 | The <unk> stocks decline was about as expected , but the Comex gain was n't . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R
367 | However , this was <unk> aside by concern over the stock market , the analyst said . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
368 | `` It was simply <unk> , '' he said , and selling by funds that are computer <unk> helped depress prices . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
369 | Futures prices eased , more in reaction to Hurricane Jerry than to any influence of the stock market . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
370 | The December contract ended with a loss of #.## cent a pound , at ##.## cents . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
371 | Prices rose sharply Friday as the storm approached Texas and Louisiana , which is part of the Mississippi Delta <unk> area . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
372 | However , after <unk> the potential effect of the hurricane , prices began to slip late Friday , Mr. Simon said . ||| SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
373 | That selling continued yesterday and kept prices under pressure , he said . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
374 | <unk> weather is being predicted for the high <unk> of Texas and the northern states of the Delta during the coming weekend , Mr. Simon said . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
375 | That has n't yet captured traders ' attention , he added . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
376 | Futures prices declined . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT REDUCE_R
377 | The March contract was off #.## cent a pound at ##.## cents . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
378 | At one point in early trading the March price rose to as high as ##.## cents when the stock market recovered , but the price then fell back . ||| SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R
379 | India recently bought ###,### tons and was expected to buy more , the analyst said . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R
380 | Another analyst thought that India may have pulled back because of the concern over the stock market . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
381 | At any rate , she added , `` India needs the sugar , so it will be in sooner or later to buy it . '' ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R
382 | <unk> <unk> : ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R
383 | The price of the <unk> contract for October delivery dropped its maximum permissible daily limit of #.# cents a pound . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
384 | The prices of most grain futures contracts rose slightly yesterday out of relief that the stock market was showing signs of <unk> . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
385 | Earlier in the session , the prices of several <unk> contracts set new lows . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R
386 | A broad rally began when several major processors began buying futures contracts , apparently to take advantage of the price dip . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
387 | The <unk> spokesman said the third-quarter earnings that the company plans to report Oct. ## are expected to be up . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
388 | <unk> ' <unk> is based in <unk> , Mass . ||| SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
389 | <unk> Operations , a food services concern , and <unk> , a holding company with interests in oil and natural gas and financial services , are based in Toronto . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R SHIFT REDUCE_R
390 | The company earned $ ## million , or ## cents a share , in the year-ago quarter . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
391 | Earlier repairs <unk> the CFCs out of the home through a <unk> directly into the <unk> . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R
392 | CFCs are widely used as <unk> , <unk> and fire <unk> . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
393 | However , <unk> said it believes the reserves in the field are about ## million barrels of oil . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
394 | <unk> , an independent oil and gas concern , is the operator and owns a ## % interest in the new field , called Northeast <unk> . ||| SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
395 | The <unk> contract area is held with <unk> , the <unk> state oil company . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
396 | But because of confusion , it took those credits again in reporting its results through the first nine months . ||| SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
397 | In <unk> Exchange trading , <unk> closed unchanged at ##.### Canadian dollars -LRB- US$ #.## -RRB- . ||| SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
398 | <unk> is a <unk> maker of telecommunications products and provider of telecommunications and computer services . ||| SHIFT SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
399 | <unk> said the agreement calls for it to make a $ ##-a-share cash tender offer for all shares outstanding of <unk> . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_R SHIFT SHIFT SHIFT REDUCE_L REDUCE_L REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT SHIFT REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
400 | <unk> said the tender offer is <unk> on , among other things , holders <unk> at least ## % of the shares outstanding , other than the shares held by Mr. <unk> . ||| SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT REDUCE_L REDUCE_R SHIFT REDUCE_R REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_R SHIFT REDUCE_L SHIFT REDUCE_L SHIFT SHIFT SHIFT REDUCE_L SHIFT REDUCE_R SHIFT REDUCE_R SHIFT SHIFT SHIFT SHIFT REDUCE_L SHIFT SHIFT SHIFT SHIFT REDUCE_L REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R REDUCE_R SHIFT REDUCE_R
401 | 


--------------------------------------------------------------------------------
/tutorial_bilstm_tagger.py:
--------------------------------------------------------------------------------
  1 | from collections import Counter, defaultdict
  2 | from itertools import count
  3 | import random
  4 | 
  5 | import dynet as dy
  6 | import numpy as np
  7 | 
  8 | # format of files: each line is "word1/tag2 word2/tag2 ..."
  9 | train_file="WSJ_TRAIN"
 10 | dev_file="WSJ_DEV"
 11 | 
 12 | class Vocab:
 13 |     def __init__(self, w2i=None):
 14 |         if w2i is None: w2i = defaultdict(count(0).next)
 15 |         self.w2i = dict(w2i)
 16 |         self.i2w = {i:w for w,i in w2i.iteritems()}
 17 |     @classmethod
 18 |     def from_corpus(cls, corpus):
 19 |         w2i = defaultdict(count(0).next)
 20 |         for sent in corpus:
 21 |             [w2i[word] for word in sent]
 22 |         return Vocab(w2i)
 23 | 
 24 |     def size(self): return len(self.w2i.keys())
 25 | 
 26 | def read(fname):
 27 |     """
 28 |     Read a POS-tagged file where each line is of the form "word1/tag2 word2/tag2 ..."
 29 |     Yields lists of the form [(word1,tag1), (word2,tag2), ...]
 30 |     """
 31 |     with file(fname) as fh:
 32 |         for line in fh:
 33 |             line = line.strip().split()
 34 |             sent = [tuple(x.rsplit("/",1)) for x in line]
 35 |             yield sent
 36 | 
 37 | train=list(read(train_file))
 38 | dev=list(read(dev_file))
 39 | words=[]
 40 | tags=[]
 41 | chars=set()
 42 | wc=Counter()
 43 | for sent in train:
 44 |     for w,p in sent:
 45 |         words.append(w)
 46 |         tags.append(p)
 47 |         chars.update(w)
 48 |         wc[w]+=1
 49 | words.append("_UNK_")
 50 | chars.add("<*>")
 51 | 
 52 | vw = Vocab.from_corpus([words]) 
 53 | vt = Vocab.from_corpus([tags])
 54 | vc = Vocab.from_corpus([chars])
 55 | UNK = vw.w2i["_UNK_"]
 56 | 
 57 | nwords = vw.size()
 58 | ntags  = vt.size()
 59 | nchars  = vc.size()
 60 | 
 61 | # DyNet Starts
 62 | 
 63 | model = dy.Model()
 64 | trainer = dy.AdamTrainer(model)
 65 | 
 66 | WORDS_LOOKUP = model.add_lookup_parameters((nwords, 128))
 67 | CHARS_LOOKUP = model.add_lookup_parameters((nchars, 20))
 68 | p_t1  = model.add_lookup_parameters((ntags, 30))
 69 | 
 70 | # MLP on top of biLSTM outputs 100 -> 32 -> ntags
 71 | pH = model.add_parameters((32, 50*2))
 72 | pO = model.add_parameters((ntags, 32))
 73 | 
 74 | # word-level LSTMs
 75 | fwdRNN = dy.LSTMBuilder(1, 128, 50, model) # layers, in-dim, out-dim, model
 76 | bwdRNN = dy.LSTMBuilder(1, 128, 50, model)
 77 | 
 78 | # char-level LSTMs
 79 | cFwdRNN = dy.LSTMBuilder(1, 20, 64, model)
 80 | cBwdRNN = dy.LSTMBuilder(1, 20, 64, model)
 81 | 
 82 | def word_rep(w, cf_init, cb_init):
 83 |     if wc[w] > 5:
 84 |         w_index = vw.w2i[w]
 85 |         return WORDS_LOOKUP[w_index]
 86 |     else:
 87 |         pad_char = vc.w2i["<*>"]
 88 |         char_ids = [pad_char] + [vc.w2i[c] for c in w] + [pad_char]
 89 |         char_embs = [CHARS_LOOKUP[cid] for cid in char_ids]
 90 |         fw_exps = cf_init.transduce(char_embs)
 91 |         bw_exps = cb_init.transduce(reversed(char_embs))
 92 |         return dy.concatenate([ fw_exps[-1], bw_exps[-1] ])
 93 | 
 94 | def build_tagging_graph(words):
 95 |     dy.renew_cg()
 96 |     # parameters -> expressions
 97 |     H = dy.parameter(pH)
 98 |     O = dy.parameter(pO)
 99 | 
100 |     # initialize the RNNs
101 |     f_init = fwdRNN.initial_state()
102 |     b_init = bwdRNN.initial_state()
103 | 
104 |     cf_init = cFwdRNN.initial_state()
105 |     cb_init = cBwdRNN.initial_state()
106 | 
107 |     # get the word vectors. word_rep(...) returns a 128-dim vector expression for each word.
108 |     wembs = [word_rep(w, cf_init, cb_init) for w in words]
109 |     wembs = [dy.noise(we,0.1) for we in wembs] # optional
110 | 
111 |     # feed word vectors into biLSTM
112 |     fw_exps = f_init.transduce(wembs)
113 |     bw_exps = b_init.transduce(reversed(wembs))
114 | # OR
115 | #    fw_exps = []
116 | #    s = f_init
117 | #    for we in wembs:
118 | #        s = s.add_input(we)
119 | #        fw_exps.append(s.output())
120 | #    bw_exps = []
121 | #    s = b_init
122 | #    for we in reversed(wembs):
123 | #        s = s.add_input(we)
124 | #        bw_exps.append(s.output())
125 | 
126 |     # biLSTM states
127 |     bi_exps = [dy.concatenate([f,b]) for f,b in zip(fw_exps, reversed(bw_exps))]
128 | 
129 |     # feed each biLSTM state to an MLP
130 |     exps = []
131 |     for x in bi_exps:
132 |         r_t = O*(dy.tanh(H * x))
133 |         exps.append(r_t)
134 | 
135 |     return exps
136 | 
137 | def sent_loss(words, tags):
138 |     vecs = build_tagging_graph(words)
139 |     errs = []
140 |     for v,t in zip(vecs,tags):
141 |         tid = vt.w2i[t]
142 |         err = dy.pickneglogsoftmax(v, tid)
143 |         errs.append(err)
144 |     return dy.esum(errs)
145 | 
146 | def tag_sent(words):
147 |     vecs = build_tagging_graph(words)
148 |     vecs = [dy.softmax(v) for v in vecs]
149 |     probs = [v.npvalue() for v in vecs]
150 |     tags = []
151 |     for prb in probs:
152 |         tag = np.argmax(prb)
153 |         tags.append(vt.i2w[tag])
154 |     return zip(words, tags)
155 | 
156 | num_tagged = cum_loss = 0
157 | for ITER in xrange(50):
158 |     random.shuffle(train)
159 |     for i,s in enumerate(train,1):
160 |         if i > 0 and i % 500 == 0:   # print status
161 |             trainer.status()
162 |             print cum_loss / num_tagged
163 |             cum_loss = num_tagged = 0
164 |             num_tagged = 0
165 |         if i % 10000 == 0 or i == len(train)-1: # eval on dev
166 |             good_sent = bad_sent = good = bad = 0.0
167 |             for sent in dev:
168 |                 words = [w for w,t in sent]
169 |                 golds = [t for w,t in sent]
170 |                 tags = [t for w,t in tag_sent(words)]
171 |                 if tags == golds: good_sent += 1
172 |                 else: bad_sent += 1
173 |                 for go,gu in zip(golds,tags):
174 |                     if go == gu: good += 1
175 |                     else: bad += 1
176 |             print good/(good+bad), good_sent/(good_sent+bad_sent)
177 |         # train on sent
178 |         words = [w for w,t in s]
179 |         golds = [t for w,t in s]
180 | 
181 |         loss_exp =  sent_loss(words, golds)
182 |         cum_loss += loss_exp.scalar_value()
183 |         num_tagged += len(golds)
184 |         loss_exp.backward()
185 |         trainer.update()
186 |     print "epoch %r finished" % ITER
187 |     trainer.update_epoch(1.0)
188 | 
189 | 
190 | 


--------------------------------------------------------------------------------
/tutorial_bilstm_viterbi_tagger.py:
--------------------------------------------------------------------------------
  1 | from collections import Counter, defaultdict
  2 | from itertools import count, chain
  3 | import random
  4 | import math
  5 | 
  6 | import dynet as dy
  7 | import numpy as np
  8 | 
  9 | # format of files: each line is "word1/tag2 word2/tag2 ..."
 10 | train_file="WSJ_TRAIN"
 11 | test_file="WSJ_DEV"
 12 | 
 13 | MAX_LIK_ITERS = 3
 14 | SMALL_NUMBER = -1e10
 15 | MARGIN = 0
 16 | 
 17 | class Vocab:
 18 |     def __init__(self, w2i=None):
 19 |         if w2i is None: w2i = defaultdict(count(0).next)
 20 |         self.w2i = dict(w2i)
 21 |         self.i2w = {i:w for w,i in w2i.iteritems()}
 22 |     @classmethod
 23 |     def from_corpus(cls, corpus):
 24 |         w2i = defaultdict(count(0).next)
 25 |         for sent in corpus:
 26 |             [w2i[word] for word in sent]
 27 |         return Vocab(w2i)
 28 | 
 29 |     def size(self): return len(self.w2i.keys())
 30 | 
 31 | def read(fname):
 32 |     """
 33 |     Read a POS-tagged file where each line is of the form "word1/tag2 word2/tag2 ..."
 34 |     Yields lists of the form [(word1,tag1), (word2,tag2), ...]
 35 |     """
 36 |     with file(fname) as fh:
 37 |         for line in fh:
 38 |             line = line.strip().split()
 39 |             sent = [tuple(x.rsplit("/",1)) for x in line]
 40 |             yield sent
 41 | 
 42 | train=list(read(train_file))
 43 | test=list(read(test_file))
 44 | words=[]
 45 | tags=[]
 46 | chars=set()
 47 | wc=Counter()
 48 | for sent in train:
 49 |     for w,p in sent:
 50 |         words.append(w)
 51 |         tags.append(p)
 52 |         chars.update(w)
 53 |         wc[w]+=1
 54 | words.append("_UNK_")
 55 | words.append("_S_")
 56 | tags.append("_S_")
 57 | chars.add("<*>")
 58 | 
 59 | vw = Vocab.from_corpus([words]) # TODO Vocab
 60 | vt = Vocab.from_corpus([tags])
 61 | vc = Vocab.from_corpus([chars])
 62 | UNK = vw.w2i["_UNK_"]
 63 | S_W = vw.w2i["_S_"]
 64 | S_T = vt.w2i["_S_"]
 65 | 
 66 | nwords = vw.size()
 67 | ntags  = vt.size()
 68 | nchars  = vc.size()
 69 | 
 70 | # DyNet Starts
 71 | 
 72 | model = dy.Model()
 73 | trainer = dy.AdamTrainer(model)
 74 | 
 75 | WORDS_LOOKUP = model.add_lookup_parameters((nwords, 128))
 76 | CHARS_LOOKUP = model.add_lookup_parameters((nchars, 20))
 77 | TRANS_LOOKUP = model.add_lookup_parameters((ntags, ntags))
 78 | p_t1  = model.add_lookup_parameters((ntags, 30))
 79 | 
 80 | # MLP on top of biLSTM outputs 100 -> 32 -> ntags
 81 | pH = model.add_parameters((32, 50*2))
 82 | pO = model.add_parameters((ntags, 32))
 83 | 
 84 | # word-level LSTMs
 85 | fwdRNN = dy.LSTMBuilder(1, 128, 50, model) # layers, in-dim, out-dim, model
 86 | bwdRNN = dy.LSTMBuilder(1, 128, 50, model)
 87 | 
 88 | # char-level LSTMs
 89 | cFwdRNN = dy.LSTMBuilder(1, 20, 64, model)
 90 | cBwdRNN = dy.LSTMBuilder(1, 20, 64, model)
 91 | 
 92 | def word_rep(w, cf_init, cb_init):
 93 |     if wc[w] > 5:
 94 |         w_index = vw.w2i[w]
 95 |         return WORDS_LOOKUP[w_index]
 96 |     else:
 97 |         pad_char = vc.w2i["<*>"]
 98 |         char_ids = [pad_char] + [vc.w2i[c] for c in w] + [pad_char]
 99 |         char_embs = [CHARS_LOOKUP[cid] for cid in char_ids]
100 |         fw_exps = cf_init.transduce(char_embs)
101 |         bw_exps = cb_init.transduce(reversed(char_embs))
102 |         return dy.concatenate([ fw_exps[-1], bw_exps[-1] ])
103 | 
104 | def build_tagging_graph(words):
105 |     dy.renew_cg()
106 |     # parameters -> expressions
107 |     H = dy.parameter(pH)
108 |     O = dy.parameter(pO)
109 | 
110 |     # initialize the RNNs
111 |     f_init = fwdRNN.initial_state()
112 |     b_init = bwdRNN.initial_state()
113 | 
114 |     cf_init = cFwdRNN.initial_state()
115 |     cb_init = cBwdRNN.initial_state()
116 | 
117 |     # get the word vectors. word_rep(...) returns a 128-dim vector expression for each word.
118 |     wembs = [word_rep(w, cf_init, cb_init) for w in words]
119 |     wembs = [dy.noise(we,0.1) for we in wembs] # optional
120 | 
121 |     # feed word vectors into biLSTM
122 |     fw_exps = f_init.transduce(wembs)
123 |     bw_exps = b_init.transduce(reversed(wembs))
124 | # OR
125 | #    fw_exps = []
126 | #    s = f_init
127 | #    for we in wembs:
128 | #        s = s.add_input(we)
129 | #        fw_exps.append(s.output())
130 | #    bw_exps = []
131 | #    s = b_init
132 | #    for we in reversed(wembs):
133 | #        s = s.add_input(we)
134 | #        bw_exps.append(s.output())
135 | 
136 |     # biLSTM states
137 |     bi_exps = [dy.concatenate([f,b]) for f,b in zip(fw_exps, reversed(bw_exps))]
138 | 
139 |     # feed each biLSTM state to an MLP
140 |     exps = []
141 |     for bi in bi_exps:
142 |         r_t = O*(dy.tanh(H * bi))
143 |         exps.append(r_t)
144 | 
145 |     return exps
146 | 
147 | def viterbi_decoding(vecs, gold_tags = []):
148 |     # Initialize
149 |     init_prob = [SMALL_NUMBER] * ntags
150 |     init_prob[S_T] = 0
151 |     for_expr = dy.inputVector(init_prob)
152 |     best_ids = []
153 |     trans_exprs = [TRANS_LOOKUP[tid] for tid in range(ntags)]
154 |     # Perform the forward pass through the sentence
155 |     for i, vec in enumerate(vecs):
156 |         my_best_ids = []
157 |         my_best_exprs = []
158 |         for next_tag in range(ntags):
159 |             # Calculate vector for single next tag
160 |             next_single_expr = for_expr + trans_exprs[next_tag]
161 |             next_single = next_single_expr.npvalue()
162 |             # Find and save the best score
163 |             my_best_id = np.argmax(next_single)
164 |             my_best_ids.append(my_best_id)
165 |             my_best_exprs.append(dy.pick(next_single_expr, my_best_id))
166 |         # Concatenate the scores for all vectors together
167 |         for_expr = dy.concatenate(my_best_exprs) + vec
168 |         # Give a bonus to all but the correct tag if using margin
169 |         if MARGIN != 0 and len(gold_tags) != 0:
170 |             adjust = [MARGIN] * ntags
171 |             adjust[vt.w2i[gold_tags[i]]] = 0
172 |             for_expr = for_expr + dy.inputVector(adjust)
173 |         # Save the best ids
174 |         best_ids.append(my_best_ids)
175 |     # Perform the final step to the sentence terminal symbol
176 |     next_single_expr = for_expr + trans_exprs[S_T]
177 |     next_single = next_single_expr.npvalue()
178 |     my_best_id = np.argmax(next_single)
179 |     best_expr = dy.pick(next_single_expr, my_best_id)
180 |     # Perform the reverse pass
181 |     best_path = [vt.i2w[my_best_id]]
182 |     for my_best_ids in reversed(best_ids):
183 |         my_best_id = my_best_ids[my_best_id]
184 |         best_path.append(vt.i2w[my_best_id])
185 |     best_path.pop() # Remove final <s>
186 |     best_path.reverse()
187 |     # Return the best path and best score as an expression
188 |     return best_path, best_expr
189 | 
190 | def forced_decoding(vecs, tags):
191 |     # Initialize
192 |     for_expr = dy.scalarInput(0)
193 |     for_tag = S_T
194 |     # Perform the forward pass through the sentence
195 |     for i, vec in enumerate(vecs): 
196 |         my_tag = vt.w2i[tags[i]]
197 |         for_expr = for_expr + dy.pick(TRANS_LOOKUP[my_tag], for_tag) + vec[my_tag]
198 |         for_tag = my_tag
199 |     for_expr = for_expr + dy.pick(TRANS_LOOKUP[S_T], for_tag)
200 |     return for_expr
201 | 
202 | def viterbi_sent_loss(words, tags):
203 |     vecs = build_tagging_graph(words)
204 |     viterbi_tags, viterbi_score = viterbi_decoding(vecs, tags)
205 |     if viterbi_tags != tags:
206 |         reference_score = forced_decoding(vecs, tags)
207 |         return viterbi_score - reference_score
208 |     else:
209 |         return dy.scalarInput(0)
210 | 
211 | def sent_loss(words, tags):
212 |     vecs = build_tagging_graph(words)
213 |     errs = []
214 |     for v,t in zip(vecs,tags):
215 |         tid = vt.w2i[t]
216 |         err = dy.pickneglogsoftmax(v, tid)
217 |         errs.append(err)
218 |     return dy.esum(errs)
219 | 
220 | def tag_sent(words):
221 |     vecs = build_tagging_graph(words)
222 |     vecs = [dy.softmax(v) for v in vecs]
223 |     probs = [v.npvalue() for v in vecs]
224 |     tags = []
225 |     for prb in probs:
226 |         tag = np.argmax(prb)
227 |         tags.append(vt.i2w[tag])
228 |     return tags
229 | 
230 | num_tagged = cum_loss = 0
231 | for ITER in xrange(50):
232 |     random.shuffle(train)
233 |     for i,s in enumerate(train,1):
234 |         if i % 500 == 0:
235 |             trainer.status()
236 |             print cum_loss / num_tagged
237 |             cum_loss = 0
238 |             num_tagged = 0
239 |         if i % 10000 == 0 or i == len(train)-1: 
240 |             good_sent = bad_sent = good = bad = 0.0
241 |             for sent in test:
242 |                 words = [w for w,t in sent]
243 |                 golds = [t for w,t in sent]
244 |                 if ITER < MAX_LIK_ITERS:
245 |                     tags = tag_sent(words)
246 |                 else:
247 |                     vecs = build_tagging_graph(words)
248 |                     tags, loss_exp = viterbi_decoding(vecs)
249 |                 if tags == golds: good_sent += 1
250 |                 else: bad_sent += 1
251 |                 for go,gu in zip(golds,tags):
252 |                     if go == gu: good += 1
253 |                     else: bad += 1
254 |             print good/(good+bad), good_sent/(good_sent+bad_sent)
255 |         # train on sent
256 |         words = [w for w,t in s]
257 |         golds = [t for w,t in s]
258 | 
259 |         if ITER < MAX_LIK_ITERS:
260 |             loss_exp =  sent_loss(words, golds)
261 |         else:
262 |             loss_exp =  viterbi_sent_loss(words, golds)
263 |         cum_loss += loss_exp.scalar_value()
264 |         num_tagged += len(golds)
265 |         loss_exp.backward()
266 |         trainer.update()
267 |     print "epoch %r finished" % ITER
268 |     trainer.update_epoch(1.0)
269 | 
270 | 
271 | 


--------------------------------------------------------------------------------
/tutorial_parser.ipynb:
--------------------------------------------------------------------------------
  1 | {
  2 |  "cells": [
  3 |   {
  4 |    "cell_type": "code",
  5 |    "execution_count": 2,
  6 |    "metadata": {
  7 |     "collapsed": true
  8 |    },
  9 |    "outputs": [],
 10 |    "source": [
 11 |     "from __future__ import division\n",
 12 |     "from operator import itemgetter\n",
 13 |     "from itertools import count\n",
 14 |     "from collections import Counter, defaultdict\n",
 15 |     "import random\n",
 16 |     "import numpy as np\n",
 17 |     "import dynet as dy\n",
 18 |     "import re"
 19 |    ]
 20 |   },
 21 |   {
 22 |    "cell_type": "code",
 23 |    "execution_count": 3,
 24 |    "metadata": {
 25 |     "collapsed": true
 26 |    },
 27 |    "outputs": [],
 28 |    "source": [
 29 |     "# config\n",
 30 |     "WORD_DIM = 64\n",
 31 |     "LSTM_DIM = 64\n",
 32 |     "ACTION_DIM = 32"
 33 |    ]
 34 |   },
 35 |   {
 36 |    "cell_type": "code",
 37 |    "execution_count": 4,
 38 |    "metadata": {
 39 |     "collapsed": true
 40 |    },
 41 |    "outputs": [],
 42 |    "source": [
 43 |     "# represents a bidirectional mapping from strings to ints\n",
 44 |     "class Vocab(object):\n",
 45 |     "    def __init__(self, w2i):\n",
 46 |     "        self.w2i = dict(w2i)\n",
 47 |     "        self.i2w = {i:w for w,i in w2i.iteritems()}\n",
 48 |     "\n",
 49 |     "    @classmethod\n",
 50 |     "    def from_list(cls, words):\n",
 51 |     "        w2i = {}\n",
 52 |     "        idx = 0\n",
 53 |     "        for word in words:\n",
 54 |     "            w2i[word] = idx\n",
 55 |     "            idx += 1\n",
 56 |     "        return Vocab(w2i)\n",
 57 |     "\n",
 58 |     "    @classmethod\n",
 59 |     "    def from_file(cls, vocab_fname):\n",
 60 |     "        words = []\n",
 61 |     "        with file(vocab_fname) as fh:\n",
 62 |     "            for line in fh:\n",
 63 |     "                line.strip()\n",
 64 |     "                word, count = line.split()\n",
 65 |     "                words.append(word)\n",
 66 |     "        return Vocab.from_list(words)\n",
 67 |     "\n",
 68 |     "    def size(self): return len(self.w2i.keys())"
 69 |    ]
 70 |   },
 71 |   {
 72 |    "cell_type": "code",
 73 |    "execution_count": 5,
 74 |    "metadata": {
 75 |     "collapsed": true
 76 |    },
 77 |    "outputs": [],
 78 |    "source": [
 79 |     "# format:\n",
 80 |     "# John left . ||| SHIFT SHIFT REDUCE_L SHIFT REDUCE_R\n",
 81 |     "def read_oracle(fname, vw, va):\n",
 82 |     "    with file(fname) as fh:\n",
 83 |     "        for line in fh:\n",
 84 |     "            line = line.strip()\n",
 85 |     "            ssent, sacts = re.split(r' \\|\\|\\| ', line)\n",
 86 |     "            sent = [vw.w2i[x] for x in ssent.split()]\n",
 87 |     "            acts = [va.w2i[x] for x in sacts.split()]\n",
 88 |     "            sent.reverse()\n",
 89 |     "            acts.reverse()\n",
 90 |     "            yield (sent, acts)"
 91 |    ]
 92 |   },
 93 |   {
 94 |    "cell_type": "code",
 95 |    "execution_count": 6,
 96 |    "metadata": {
 97 |     "collapsed": true
 98 |    },
 99 |    "outputs": [],
100 |    "source": [
101 |     "class StackRNN(object):\n",
102 |     "    def __init__(self, rnn, p_empty_embedding = None):\n",
103 |     "        self.s = [(rnn.initial_state(), None)]\n",
104 |     "        self.empty = None\n",
105 |     "        if p_empty_embedding:\n",
106 |     "            self.empty = dy.parameter(p_empty_embedding)\n",
107 |     "    def push(self, expr, extra=None):\n",
108 |     "        self.s.append((self.s[-1][0].add_input(expr), extra))\n",
109 |     "    def pop(self):\n",
110 |     "        return self.s.pop()[1] # return \"extra\" (i.e., whatever the caller wants or None)\n",
111 |     "    def embedding(self):\n",
112 |     "        # work around since inital_state.output() is None\n",
113 |     "        return self.s[-1][0].output() if len(self.s) > 1 else self.empty\n",
114 |     "    def __len__(self):\n",
115 |     "        return len(self.s) - 1"
116 |    ]
117 |   },
118 |   {
119 |    "cell_type": "code",
120 |    "execution_count": 7,
121 |    "metadata": {
122 |     "collapsed": false
123 |    },
124 |    "outputs": [],
125 |    "source": [
126 |     "# actions the parser can take\n",
127 |     "acts = ['SHIFT', 'REDUCE_L', 'REDUCE_R']\n",
128 |     "vocab_acts = Vocab.from_list(acts)\n",
129 |     "SHIFT = vocab_acts.w2i['SHIFT']\n",
130 |     "REDUCE_L = vocab_acts.w2i['REDUCE_L']\n",
131 |     "REDUCE_R = vocab_acts.w2i['REDUCE_R']\n",
132 |     "NUM_ACTIONS = vocab_acts.size()\n",
133 |     "\n",
134 |     "class TransitionParser(object):\n",
135 |     "    def __init__(self, model, vocab):\n",
136 |     "        self.vocab = vocab\n",
137 |     "        # syntactic composition\n",
138 |     "        self.pW_comp = model.add_parameters((LSTM_DIM, LSTM_DIM * 2))\n",
139 |     "        self.pb_comp = model.add_parameters(LSTM_DIM)\n",
140 |     "        # parser state to hidden\n",
141 |     "        self.pW_s2h = model.add_parameters((LSTM_DIM, LSTM_DIM * 2))\n",
142 |     "        self.pb_s2h = model.add_parameters(LSTM_DIM)\n",
143 |     "        # hidden to action\n",
144 |     "        self.pW_act = model.add_parameters((NUM_ACTIONS, LSTM_DIM))\n",
145 |     "        self.pb_act = model.add_parameters(NUM_ACTIONS)\n",
146 |     "\n",
147 |     "        # layers, in-dim, out-dim, model\n",
148 |     "        self.buffRNN = dy.LSTMBuilder(1, WORD_DIM, LSTM_DIM, model)\n",
149 |     "        self.stackRNN = dy.LSTMBuilder(1, WORD_DIM, LSTM_DIM, model)\n",
150 |     "        self.pempty_buffer_emb = model.add_parameters(LSTM_DIM)\n",
151 |     "        self.WORDS_LOOKUP = model.add_lookup_parameters((vocab.size(), WORD_DIM))\n",
152 |     "\n",
153 |     "    # Returns an expression of the loss for the sequence of actions.\n",
154 |     "    # (that is, the oracle_actions if present or the predicted sequence otherwise)\n",
155 |     "    def parse(self, tokens, oracle_actions=None):\n",
156 |     "        def _valid_actions(stack, buffer):\n",
157 |     "            valid_actions = []\n",
158 |     "            if len(buffer) > 0:\n",
159 |     "                valid_actions += [SHIFT]\n",
160 |     "            if len(stack) >= 2:\n",
161 |     "                valid_actions += [REDUCE_L, REDUCE_R]\n",
162 |     "            return valid_actions\n",
163 |     "\n",
164 |     "        dy.renew_cg() # each sentence gets its own graph\n",
165 |     "        if oracle_actions: oracle_actions = list(oracle_actions)\n",
166 |     "        buffer = StackRNN(self.buffRNN, self.pempty_buffer_emb)\n",
167 |     "        stack = StackRNN(self.stackRNN)\n",
168 |     "    \n",
169 |     "        # Put the parameters in the cg\n",
170 |     "        W_comp = dy.parameter(self.pW_comp) # syntactic composition\n",
171 |     "        b_comp = dy.parameter(self.pb_comp)\n",
172 |     "        W_s2h = dy.parameter(self.pW_s2h)   # state to hidden\n",
173 |     "        b_s2h = dy.parameter(self.pb_s2h)\n",
174 |     "        W_act = dy.parameter(self.pW_act)   # hidden to action\n",
175 |     "        b_act = dy.parameter(self.pb_act)\n",
176 |     "    \n",
177 |     "        # We will keep track of all the losses we accumulate during parsing.\n",
178 |     "        # If some decision is unambiguous because it's the only thing valid given\n",
179 |     "        # the parser state, we will not model it. We only model what is ambiguous.\n",
180 |     "        losses = []\n",
181 |     "        \n",
182 |     "        # push the tokens onto the buffer (tokens is in reverse order)\n",
183 |     "        for tok in tokens:\n",
184 |     "            tok_embedding = self.WORDS_LOOKUP[tok]\n",
185 |     "            buffer.push(tok_embedding, (tok_embedding, self.vocab.i2w[tok]))\n",
186 |     "\n",
187 |     "        while not (len(stack) == 1 and len(buffer) == 0):\n",
188 |     "            # compute probability of each of the actions and choose an action\n",
189 |     "            # either from the oracle or if there is no oracle, based on the model\n",
190 |     "            valid_actions = _valid_actions(stack, buffer)\n",
191 |     "            log_probs = None\n",
192 |     "            action = valid_actions[0]\n",
193 |     "            if len(valid_actions) > 1:\n",
194 |     "                p_t = dy.concatenate([buffer.embedding(), stack.embedding()])\n",
195 |     "                h = dy.tanh(W_s2h * p_t + b_s2h)\n",
196 |     "                logits = W_act * h + b_act\n",
197 |     "                log_probs = dy.log_softmax(logits, valid_actions)\n",
198 |     "                if oracle_actions is None:\n",
199 |     "                    action = np.argmax(log_probs.npvalue())\n",
200 |     "            if oracle_actions is not None:\n",
201 |     "                action = oracle_actions.pop()\n",
202 |     "            if log_probs is not None:\n",
203 |     "                # append the action-specific loss\n",
204 |     "                losses.append(dy.pick(log_probs, action))\n",
205 |     "\n",
206 |     "            # execute the action to update the parser state\n",
207 |     "            if action == SHIFT:\n",
208 |     "                tok_embedding, token = buffer.pop()\n",
209 |     "                stack.push(tok_embedding, (tok_embedding, token))\n",
210 |     "            else: # one of the REDUCE actions\n",
211 |     "                right = stack.pop() # pop a stack state\n",
212 |     "                left = stack.pop()  # pop another stack state\n",
213 |     "                # figure out which is the head and which is the modifier\n",
214 |     "                head, modifier = (left, right) if action == REDUCE_R else (right, left)\n",
215 |     "        \n",
216 |     "                # compute composed representation\n",
217 |     "                head_rep, head_tok = head\n",
218 |     "                mod_rep, mod_tok = modifier\n",
219 |     "                composed_rep = dy.tanh(W_comp * dy.concatenate([head_rep, mod_rep]) + b_comp)\n",
220 |     "                \n",
221 |     "                stack.push(composed_rep, (composed_rep, head_tok))\n",
222 |     "                if oracle_actions is None:\n",
223 |     "                    print('{0} --> {1}'.format(head_tok, mod_tok))\n",
224 |     "\n",
225 |     "        # the head of the tree that remains at the top of the stack is the root\n",
226 |     "        if oracle_actions is None:\n",
227 |     "            head = stack.pop()[1]\n",
228 |     "            print('ROOT --> {0}'.format(head))\n",
229 |     "        return -dy.esum(losses) if losses else None"
230 |    ]
231 |   },
232 |   {
233 |    "cell_type": "code",
234 |    "execution_count": 8,
235 |    "metadata": {
236 |     "collapsed": true
237 |    },
238 |    "outputs": [],
239 |    "source": [
240 |     "# load training and dev data\n",
241 |     "vocab_words = Vocab.from_file('data/vocab.txt')\n",
242 |     "train = list(read_oracle('data/small-train.unk.txt', vocab_words, vocab_acts))\n",
243 |     "dev = list(read_oracle('data/small-dev.unk.txt', vocab_words, vocab_acts))\n",
244 |     "\n",
245 |     "model = dy.Model()\n",
246 |     "trainer = dy.AdamTrainer(model)\n",
247 |     "\n",
248 |     "tp = TransitionParser(model, vocab_words)"
249 |    ]
250 |   },
251 |   {
252 |    "cell_type": "code",
253 |    "execution_count": 9,
254 |    "metadata": {
255 |     "collapsed": false
256 |    },
257 |    "outputs": [
258 |     {
259 |      "name": "stdout",
260 |      "output_type": "stream",
261 |      "text": [
262 |       "[validation] epoch 0.0: per-word loss: 2.35509942849\n",
263 |       "epoch 0.01: per-word loss: 1.48255809603\n",
264 |       "epoch 0.02: per-word loss: 1.08090971986\n",
265 |       "epoch 0.03: per-word loss: 0.944152983054\n",
266 |       "epoch 0.04: per-word loss: 0.953884831015\n",
267 |       "epoch 0.05: per-word loss: 0.834949359487\n",
268 |       "epoch 0.06: per-word loss: 0.811419209329\n",
269 |       "epoch 0.07: per-word loss: 0.739562879959\n",
270 |       "epoch 0.08: per-word loss: 0.72608875857\n",
271 |       "epoch 0.09: per-word loss: 0.715399681898\n"
272 |      ]
273 |     },
274 |     {
275 |      "ename": "KeyboardInterrupt",
276 |      "evalue": "",
277 |      "output_type": "error",
278 |      "traceback": [
279 |       "\u001b[0;31m---------------------------------------------------------------------------\u001b[0m",
280 |       "\u001b[0;31mKeyboardInterrupt\u001b[0m                         Traceback (most recent call last)",
281 |       "\u001b[0;32m<ipython-input-9-5eacd1fbaf60>\u001b[0m in \u001b[0;36m<module>\u001b[0;34m()\u001b[0m\n\u001b[1;32m     15\u001b[0m                 \u001b[0mdev_words\u001b[0m \u001b[0;34m+=\u001b[0m \u001b[0mlen\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mds\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     16\u001b[0m                 \u001b[0;32mif\u001b[0m \u001b[0mloss\u001b[0m \u001b[0;32mis\u001b[0m \u001b[0;32mnot\u001b[0m \u001b[0mNone\u001b[0m\u001b[0;34m:\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0;32m---> 17\u001b[0;31m                     \u001b[0mdev_loss\u001b[0m \u001b[0;34m+=\u001b[0m \u001b[0mloss\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mscalar_value\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[0m\u001b[1;32m     18\u001b[0m             \u001b[0;32mprint\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0;34m'[validation] epoch {}: per-word loss: {}'\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mformat\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0me\u001b[0m\u001b[0;34m,\u001b[0m \u001b[0mdev_loss\u001b[0m \u001b[0;34m/\u001b[0m \u001b[0mdev_words\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n\u001b[1;32m     19\u001b[0m             \u001b[0mvalidation_losses\u001b[0m\u001b[0;34m.\u001b[0m\u001b[0mappend\u001b[0m\u001b[0;34m(\u001b[0m\u001b[0mdev_loss\u001b[0m\u001b[0;34m)\u001b[0m\u001b[0;34m\u001b[0m\u001b[0m\n",
282 |       "\u001b[0;31mKeyboardInterrupt\u001b[0m: "
283 |      ]
284 |     }
285 |    ],
286 |    "source": [
287 |     "instances_processed = 0\n",
288 |     "validation_losses = []\n",
289 |     "for epoch in range(5):\n",
290 |     "    random.shuffle(train)\n",
291 |     "    words = 0\n",
292 |     "    total_loss = 0.0\n",
293 |     "    for (s,a) in train:\n",
294 |     "        # periodically report validation loss\n",
295 |     "        e = instances_processed / len(train)\n",
296 |     "        if instances_processed % 1000 == 0:\n",
297 |     "            dev_words = 0\n",
298 |     "            dev_loss = 0.0\n",
299 |     "            for (ds, da) in dev:\n",
300 |     "                loss = tp.parse(ds, da)\n",
301 |     "                dev_words += len(ds)\n",
302 |     "                if loss is not None:\n",
303 |     "                    dev_loss += loss.scalar_value()\n",
304 |     "            print('[validation] epoch {}: per-word loss: {}'.format(e, dev_loss / dev_words))\n",
305 |     "            validation_losses.append(dev_loss)\n",
306 |     "\n",
307 |     "        # report training loss\n",
308 |     "        if instances_processed % 100 == 0 and words > 0:\n",
309 |     "            print('epoch {}: per-word loss: {}'.format(e, total_loss / words))\n",
310 |     "            words = 0\n",
311 |     "            total_loss = 0.0\n",
312 |     "    \n",
313 |     "        # here we do training\n",
314 |     "        loss = tp.parse(s, a) # returns None for 1-word sentencs (it's clear how to parse them)\n",
315 |     "        words += len(s)\n",
316 |     "        instances_processed += 1\n",
317 |     "        if loss is not None:\n",
318 |     "            total_loss += loss.scalar_value()\n",
319 |     "            loss.backward()\n",
320 |     "            trainer.update()\n"
321 |    ]
322 |   },
323 |   {
324 |    "cell_type": "code",
325 |    "execution_count": 10,
326 |    "metadata": {
327 |     "collapsed": false
328 |    },
329 |    "outputs": [
330 |     {
331 |      "name": "stdout",
332 |      "output_type": "stream",
333 |      "text": [
334 |       "fun --> is\n",
335 |       "fun --> Austin\n",
336 |       "in --> fun\n",
337 |       "<unk> --> in\n",
338 |       "<unk> --> .\n",
339 |       "ROOT --> <unk>\n"
340 |      ]
341 |     },
342 |     {
343 |      "data": {
344 |       "text/plain": [
345 |        "exprssion 316/1496"
346 |       ]
347 |      },
348 |      "execution_count": 10,
349 |      "metadata": {},
350 |      "output_type": "execute_result"
351 |     }
352 |    ],
353 |    "source": [
354 |     "s = 'Parsing in Austin is fun .'\n",
355 |     "UNK = vocab_words.w2i['<unk>']\n",
356 |     "toks = [vocab_words.w2i[x] if x in vocab_words.w2i else UNK for x in s.split()]\n",
357 |     "toks.reverse()\n",
358 |     "tp.parse(toks)"
359 |    ]
360 |   },
361 |   {
362 |    "cell_type": "code",
363 |    "execution_count": null,
364 |    "metadata": {
365 |     "collapsed": true
366 |    },
367 |    "outputs": [],
368 |    "source": []
369 |   },
370 |   {
371 |    "cell_type": "code",
372 |    "execution_count": null,
373 |    "metadata": {
374 |     "collapsed": true
375 |    },
376 |    "outputs": [],
377 |    "source": []
378 |   }
379 |  ],
380 |  "metadata": {
381 |   "kernelspec": {
382 |    "display_name": "Python 2",
383 |    "language": "python",
384 |    "name": "python2"
385 |   },
386 |   "language_info": {
387 |    "codemirror_mode": {
388 |     "name": "ipython",
389 |     "version": 2
390 |    },
391 |    "file_extension": ".py",
392 |    "mimetype": "text/x-python",
393 |    "name": "python",
394 |    "nbconvert_exporter": "python",
395 |    "pygments_lexer": "ipython2",
396 |    "version": "2.7.10"
397 |   }
398 |  },
399 |  "nbformat": 4,
400 |  "nbformat_minor": 0
401 | }
402 | 


--------------------------------------------------------------------------------
/tutorial_rnnlm.py:
--------------------------------------------------------------------------------
  1 | from collections import Counter, defaultdict
  2 | from itertools import count
  3 | import random
  4 | 
  5 | import dynet as dy
  6 | import numpy as np
  7 | 
  8 | # format of files: each line is "word1/tag2 word2/tag2 ..."
  9 | train_file="CHAR_TRAIN"
 10 | test_file="CHAR_DEV"
 11 | 
 12 | class Vocab:
 13 |     def __init__(self, w2i=None):
 14 |         if w2i is None: w2i = defaultdict(count(0).next)
 15 |         self.w2i = dict(w2i)
 16 |         self.i2w = {i:w for w,i in w2i.iteritems()}
 17 |     @classmethod
 18 |     def from_corpus(cls, corpus):
 19 |         w2i = defaultdict(count(0).next)
 20 |         for sent in corpus:
 21 |             [w2i[word] for word in sent]
 22 |         return Vocab(w2i)
 23 | 
 24 |     def size(self): return len(self.w2i.keys())
 25 | 
 26 | def read(fname):
 27 |     """
 28 |     Read a file where each line is of the form "word1 word2 ..."
 29 |     Yields lists of the form [word1, word2, ...]
 30 |     """
 31 |     with file(fname) as fh:
 32 |         for line in fh:
 33 |             sent = line.strip().split()
 34 |             sent.append("<s>")
 35 |             yield sent
 36 | 
 37 | train=list(read(train_file))
 38 | test=list(read(test_file))
 39 | words=[]
 40 | wc=Counter()
 41 | for sent in train:
 42 |     for w in sent:
 43 |         words.append(w)
 44 |         wc[w]+=1
 45 | 
 46 | vw = Vocab.from_corpus([words])
 47 | S = vw.w2i["<s>"]
 48 | 
 49 | nwords = vw.size()
 50 | 
 51 | # DyNet Starts
 52 | 
 53 | model = dy.Model()
 54 | trainer = dy.AdamTrainer(model)
 55 | 
 56 | # Lookup parameters for word embeddings
 57 | WORDS_LOOKUP = model.add_lookup_parameters((nwords, 64))
 58 | 
 59 | # Word-level LSTM (layers=1, input=64, output=128, model)
 60 | RNN = dy.LSTMBuilder(1, 64, 128, model)
 61 | 
 62 | # Softmax weights/biases on top of LSTM outputs
 63 | W_sm = model.add_parameters((nwords, 128))
 64 | b_sm = model.add_parameters(nwords)
 65 | 
 66 | # Build the language model graph
 67 | def calc_lm_loss(sent):
 68 | 
 69 |     dy.renew_cg()
 70 |     # parameters -> expressions
 71 |     W_exp = dy.parameter(W_sm)
 72 |     b_exp = dy.parameter(b_sm)
 73 | 
 74 |     # initialize the RNN
 75 |     f_init = RNN.initial_state()
 76 | 
 77 |     # get the word ids
 78 |     wids = [vw.w2i[w] for w in sent]
 79 | 
 80 |     # start the rnn by inputting "<s>"
 81 |     s = f_init.add_input(WORDS_LOOKUP[wids[-1]]) 
 82 | 
 83 |     # feed word vectors into the RNN and predict the next word
 84 |     losses = []
 85 |     for wid in wids:
 86 |         # calculate the softmax and loss
 87 |         score = W_exp * s.output() + b_exp
 88 |         loss = dy.pickneglogsoftmax(score, wid)
 89 |         losses.append(loss)
 90 |         # update the state of the RNN
 91 |         s = s.add_input(WORDS_LOOKUP[wid]) 
 92 |     
 93 |     return dy.esum(losses)
 94 | 
 95 | num_tagged = cum_loss = 0
 96 | for ITER in xrange(50):
 97 |     random.shuffle(train)
 98 |     for i,s in enumerate(train,1):
 99 |         if i % 500 == 0:
100 |             trainer.status()
101 |             print cum_loss / num_tagged
102 |             cum_loss = 0
103 |             num_tagged = 0
104 |         if i % 10000 == 0 or i == len(train)-1:
105 |             dev_loss = dev_words = 0
106 |             for sent in test:
107 |                 loss_exp = calc_lm_loss(sent)
108 |                 dev_loss += loss_exp.scalar_value()
109 |                 dev_words += len(sent)
110 |             print dev_loss / dev_words
111 |         # train on sent
112 | 
113 |         loss_exp = calc_lm_loss(s)
114 |         cum_loss += loss_exp.scalar_value()
115 |         num_tagged += len(s)
116 |         loss_exp.backward()
117 |         trainer.update()
118 |     print "epoch %r finished" % ITER
119 |     trainer.update_epoch(1.0)
120 | 
121 | 
122 | 


--------------------------------------------------------------------------------
/tutorial_rnnlm_minibatch.py:
--------------------------------------------------------------------------------
  1 | from collections import Counter, defaultdict
  2 | from itertools import count
  3 | import random
  4 | 
  5 | import dynet as dy
  6 | import numpy as np
  7 | 
  8 | # format of files: each line is "word1/tag2 word2/tag2 ..."
  9 | train_file="CHAR_TRAIN"
 10 | test_file="CHAR_DEV"
 11 | 
 12 | MB_SIZE = 10
 13 | 
 14 | class Vocab:
 15 |     def __init__(self, w2i=None):
 16 |         if w2i is None: w2i = defaultdict(count(0).next)
 17 |         self.w2i = dict(w2i)
 18 |         self.i2w = {i:w for w,i in w2i.iteritems()}
 19 |     @classmethod
 20 |     def from_corpus(cls, corpus):
 21 |         w2i = defaultdict(count(0).next)
 22 |         for sent in corpus:
 23 |             [w2i[word] for word in sent]
 24 |         return Vocab(w2i)
 25 | 
 26 |     def size(self): return len(self.w2i.keys())
 27 | 
 28 | def read(fname):
 29 |     """
 30 |     Read a file where each line is of the form "word1 word2 ..."
 31 |     Yields lists of the form [word1, word2, ...]
 32 |     """
 33 |     with file(fname) as fh:
 34 |         for line in fh:
 35 |             sent = line.strip().split()
 36 |             sent.append("<s>")
 37 |             yield sent
 38 | 
 39 | train=list(read(train_file))
 40 | test=list(read(test_file))
 41 | words=[]
 42 | wc=Counter()
 43 | for sent in train:
 44 |     for w in sent:
 45 |         words.append(w)
 46 |         wc[w]+=1
 47 | 
 48 | vw = Vocab.from_corpus([words])
 49 | S = vw.w2i["<s>"]
 50 | 
 51 | nwords = vw.size()
 52 | 
 53 | # DyNet Starts
 54 | 
 55 | model = dy.Model()
 56 | trainer = dy.AdamTrainer(model)
 57 | 
 58 | # Lookup parameters for word embeddings
 59 | WORDS_LOOKUP = model.add_lookup_parameters((nwords, 64))
 60 | 
 61 | # Word-level LSTM (layers=1, input=64, output=128, model)
 62 | RNN = dy.LSTMBuilder(1, 64, 128, model)
 63 | 
 64 | # Softmax weights/biases on top of LSTM outputs
 65 | W_sm = model.add_parameters((nwords, 128))
 66 | b_sm = model.add_parameters(nwords)
 67 | 
 68 | # Build the language model graph
 69 | def calc_lm_loss(sents):
 70 | 
 71 |     dy.renew_cg()
 72 |     # parameters -> expressions
 73 |     W_exp = dy.parameter(W_sm)
 74 |     b_exp = dy.parameter(b_sm)
 75 | 
 76 |     # initialize the RNN
 77 |     f_init = RNN.initial_state()
 78 | 
 79 |     # get the wids and masks for each step
 80 |     tot_words = 0
 81 |     wids = []
 82 |     masks = []
 83 |     for i in range(len(sents[0])):
 84 |         wids.append([
 85 |           (vw.w2i[sent[i]] if len(sent)>i else S) for sent in sents])
 86 |         mask = [(1 if len(sent)>i else 0) for sent in sents]
 87 |         masks.append(mask)
 88 |         tot_words += sum(mask)
 89 |     
 90 |     # start the rnn by inputting "<s>"
 91 |     init_ids = [S] * len(sents)
 92 |     s = f_init.add_input(dy.lookup_batch(WORDS_LOOKUP,init_ids))
 93 | 
 94 |     # feed word vectors into the RNN and predict the next word    
 95 |     losses = []
 96 |     for wid, mask in zip(wids, masks):
 97 |         # calculate the softmax and loss      
 98 |         score = W_exp * s.output() + b_exp
 99 |         loss = dy.pickneglogsoftmax_batch(score, wid)
100 |         # mask the loss if at least one sentence is shorter
101 |         if mask[-1] != 1:
102 |             mask_expr = dy.inputVector(mask)
103 |             mask_expr = dy.reshape(mask_expr, (1,), MB_SIZE)
104 |             loss = loss * mask_expr
105 |         losses.append(loss)
106 |         # update the state of the RNN        
107 |         wemb = dy.lookup_batch(WORDS_LOOKUP, wid)
108 |         s = s.add_input(wemb) 
109 |     
110 |     return dy.sum_batches(dy.esum(losses)), tot_words
111 | 
112 | num_tagged = cum_loss = 0
113 | # Sort training sentences in descending order and count minibatches
114 | train.sort(key=lambda x: -len(x))
115 | test.sort(key=lambda x: -len(x))
116 | train_order = [x*MB_SIZE for x in range(len(train)/MB_SIZE)]
117 | test_order = [x*MB_SIZE for x in range(len(test)/MB_SIZE)]
118 | # Perform training
119 | for ITER in xrange(50):
120 |     random.shuffle(train_order)
121 |     for i,sid in enumerate(train_order,1): 
122 |         if i % (500/MB_SIZE) == 0:
123 |             trainer.status()
124 |             print cum_loss / num_tagged
125 |             num_tagged = cum_loss = 0
126 |         if i % (10000/MB_SIZE) == 0 or i == len(train_order)-1:
127 |             dev_loss = dev_words = 0
128 |             for sid in test_order:
129 |                 loss_exp, mb_words = calc_lm_loss(test[sid:sid+MB_SIZE])
130 |                 dev_loss += loss_exp.scalar_value()
131 |                 dev_words += mb_words
132 |             print dev_loss / dev_words
133 |         # train on the minibatch
134 |         loss_exp, mb_words = calc_lm_loss(train[sid:sid+MB_SIZE])
135 |         cum_loss += loss_exp.scalar_value()
136 |         num_tagged += mb_words
137 |         loss_exp.backward()
138 |         trainer.update()
139 |     print "epoch %r finished" % ITER
140 |     trainer.update_epoch(1.0)
141 | 
142 | 
143 | 


--------------------------------------------------------------------------------
/tutorial_transition_parser.py:
--------------------------------------------------------------------------------
  1 | from __future__ import print_function
  2 | from operator import itemgetter
  3 | from itertools import count
  4 | from collections import Counter, defaultdict
  5 | import random
  6 | import dynet as dy
  7 | import numpy as np
  8 | import re
  9 | 
 10 | # actions the parser can take
 11 | SHIFT = 0
 12 | REDUCE_L = 1
 13 | REDUCE_R = 2
 14 | NUM_ACTIONS = 3
 15 | 
 16 | class Vocab:
 17 |   def __init__(self, w2i):
 18 |     self.w2i = dict(w2i)
 19 |     self.i2w = {i:w for w,i in w2i.iteritems()}
 20 |   @classmethod
 21 |   def from_list(cls, words):
 22 |     w2i = {}
 23 |     idx = 0
 24 |     for word in words:
 25 |       w2i[word] = idx
 26 |       idx += 1
 27 |     return Vocab(w2i)
 28 |   @classmethod
 29 |   def from_file(cls, vocab_fname):
 30 |     words = []
 31 |     with file(vocab_fname) as fh:
 32 |       for line in fh:
 33 |         line.strip()
 34 |         word, count = line.split()
 35 |         words.append(word)
 36 |     return Vocab.from_list(words)
 37 | 
 38 |   def size(self): return len(self.w2i.keys())
 39 | 
 40 | def read_oracle(fname, vw, va):
 41 |   with file(fname) as fh:
 42 |     for line in fh:
 43 |       line = line.strip()
 44 |       ssent, sacts = re.split(r' \|\|\| ', line)
 45 |       sent = [vw.w2i[x] for x in ssent.split()]
 46 |       acts = [va.w2i[x] for x in sacts.split()]
 47 |       yield (sent, acts)
 48 | 
 49 | WORD_DIM = 64
 50 | LSTM_DIM = 64
 51 | ACTION_DIM = 32
 52 | 
 53 | class TransitionParser:
 54 |   def __init__(self, model, vocab):
 55 |     self.vocab = vocab
 56 |     self.pW_comp = model.add_parameters((LSTM_DIM, LSTM_DIM * 2))
 57 |     self.pb_comp = model.add_parameters((LSTM_DIM, ))
 58 |     self.pW_s2h = model.add_parameters((LSTM_DIM, LSTM_DIM * 2))
 59 |     self.pb_s2h = model.add_parameters((LSTM_DIM, ))
 60 |     self.pW_act = model.add_parameters((NUM_ACTIONS, LSTM_DIM))
 61 |     self.pb_act = model.add_parameters((NUM_ACTIONS, ))
 62 | 
 63 |     # layers, in-dim, out-dim, model
 64 |     self.buffRNN = dy.LSTMBuilder(1, WORD_DIM, LSTM_DIM, model)
 65 |     self.stackRNN = dy.LSTMBuilder(1, WORD_DIM, LSTM_DIM, model)
 66 |     self.pempty_buffer_emb = model.add_parameters((LSTM_DIM,))
 67 |     nwords=vocab.size()
 68 |     self.WORDS_LOOKUP = model.add_lookup_parameters((nwords, WORD_DIM))
 69 | 
 70 |   # returns an expression of the loss for the sequence of actions
 71 |   # (that is, the oracle_actions if present or the predicted sequence otherwise)
 72 |   def parse(self, t, oracle_actions=None):
 73 |     dy.renew_cg()
 74 |     if oracle_actions:
 75 |       oracle_actions = list(oracle_actions)
 76 |       oracle_actions.reverse()
 77 |     stack_top = self.stackRNN.initial_state()
 78 |     toks = list(t)
 79 |     toks.reverse()
 80 |     stack = []
 81 |     cur = self.buffRNN.initial_state()
 82 |     buffer = []
 83 |     empty_buffer_emb = dy.parameter(self.pempty_buffer_emb)
 84 |     W_comp = dy.parameter(self.pW_comp)
 85 |     b_comp = dy.parameter(self.pb_comp)
 86 |     W_s2h = dy.parameter(self.pW_s2h)
 87 |     b_s2h = dy.parameter(self.pb_s2h)
 88 |     W_act = dy.parameter(self.pW_act)
 89 |     b_act = dy.parameter(self.pb_act)
 90 |     losses = []
 91 |     for tok in toks:
 92 |       tok_embedding = self.WORDS_LOOKUP[tok]
 93 |       cur = cur.add_input(tok_embedding)
 94 |       buffer.append((cur.output(), tok_embedding, self.vocab.i2w[tok]))
 95 | 
 96 |     while not (len(stack) == 1 and len(buffer) == 0):
 97 |       # based on parser state, get valid actions
 98 |       valid_actions = []
 99 |       if len(buffer) > 0:  # can only reduce if elements in buffer
100 |         valid_actions += [SHIFT]
101 |       if len(stack) >= 2:  # can only shift if 2 elements on stack
102 |         valid_actions += [REDUCE_L, REDUCE_R]
103 | 
104 |       # compute probability of each of the actions and choose an action
105 |       # either from the oracle or if there is no oracle, based on the model
106 |       action = valid_actions[0]
107 |       log_probs = None
108 |       if len(valid_actions) > 1:
109 |         buffer_embedding = buffer[-1][0] if buffer else empty_buffer_emb
110 |         stack_embedding = stack[-1][0].output() # the stack has something here
111 |         parser_state = dy.concatenate([buffer_embedding, stack_embedding])
112 |         h = dy.tanh(W_s2h * parser_state + b_s2h)
113 |         logits = W_act * h + b_act
114 |         log_probs = dy.log_softmax(logits, valid_actions)
115 |         if oracle_actions is None:
116 |           action = max(enumerate(log_probs.vec_value()), key=itemgetter(1))[0]
117 |       if oracle_actions is not None:
118 |         action = oracle_actions.pop()
119 |         if log_probs is not None:
120 |           # append the action-specific loss
121 |           losses.append(dy.pick(log_probs, action))
122 | 
123 |       # execute the action to update the parser state
124 |       if action == SHIFT:
125 |         _, tok_embedding, token = buffer.pop()
126 |         stack_state, _ = stack[-1] if stack else (stack_top, '<TOP>')
127 |         stack_state = stack_state.add_input(tok_embedding)
128 |         stack.append((stack_state, token))
129 |       else: # one of the reduce actions
130 |         right = stack.pop()
131 |         left = stack.pop()
132 |         head, modifier = (left, right) if action == REDUCE_R else (right, left)
133 |         top_stack_state, _ = stack[-1] if stack else (stack_top, '<TOP>')
134 |         head_rep, head_tok = head[0].output(), head[1]
135 |         mod_rep, mod_tok = modifier[0].output(), modifier[1]
136 |         composed_rep = dy.rectify(W_comp * dy.concatenate([head_rep, mod_rep]) + b_comp)
137 |         top_stack_state = top_stack_state.add_input(composed_rep)
138 |         stack.append((top_stack_state, head_tok))
139 |         if oracle_actions is None:
140 |           print('{0} --> {1}'.format(head_tok, mod_tok))
141 | 
142 |     # the head of the tree that remains at the top of the stack is now the root
143 |     if oracle_actions is None:
144 |       head = stack.pop()[1]
145 |       print('ROOT --> {0}'.format(head))
146 |     return -dy.esum(losses) if losses else None
147 | 
148 | acts = ['SHIFT', 'REDUCE_L', 'REDUCE_R']
149 | vocab_acts = Vocab.from_list(acts)
150 | 
151 | vocab_words = Vocab.from_file('data/vocab.txt')
152 | train = list(read_oracle('data/small-train.unk.txt', vocab_words, vocab_acts))
153 | dev = list(read_oracle('data/small-dev.unk.txt', vocab_words, vocab_acts))
154 | 
155 | model = dy.Model()
156 | trainer = dy.AdamTrainer(model)
157 | 
158 | tp = TransitionParser(model, vocab_words)
159 | 
160 | i = 0
161 | for epoch in range(5):
162 |   words = 0
163 |   total_loss = 0.0
164 |   for (s,a) in train:
165 |     loss = tp.parse(s, a)
166 |     words += len(s)
167 |     if loss is not None:
168 |       total_loss += loss.scalar_value()
169 |       loss.backward()
170 |       trainer.update()
171 |     e = float(i) / len(train)
172 |     if i % 50 == 0:
173 |       print('epoch {}: per-word loss: {}'.format(e, total_loss / words))
174 |       words = 0
175 |       total_loss = 0.0
176 |     if i % 500 == 0:
177 |       tp.parse(dev[209][0])
178 |       dev_words = 0
179 |       dev_loss = 0.0
180 |       for (ds, da) in dev:
181 |         loss = tp.parse(ds, da)
182 |         dev_words += len(ds)
183 |         if loss is not None:
184 |           dev_loss += loss.scalar_value()
185 |       print('[validation] epoch {}: per-word loss: {}'.format(e, dev_loss / dev_words))
186 |     i += 1
187 | 


--------------------------------------------------------------------------------