├── README.md ├── ref ├── A Conditional Random Field Word Segmenter.pdf ├── A Maximum Entropy Approach to Chinese Word Segmentation.pdf └── [2007,黄昌宁，赵海]中文分词十年回顾.pdf ├── score ├── bmm_result │ ├── pku_bmm_result.utf8 │ └── pku_bmm_score.utf8 ├── fmm_result │ ├── pku_fmm_result.utf8 │ └── pku_fmm_score.utf8 ├── pku_test_gold.utf8 ├── pku_training.utf8 └── score └── src ├── LICENSE ├── ReadMe ├── autoScore.py ├── cwsBMM.py ├── cwsBMM_NRule.py ├── cwsFMM.py ├── cwsFMM_NRule.py ├── cwsMaxEn-13f.py ├── cwsMaxEn-16f.py ├── pku_test.utf8 ├── pku_test_gold.utf8 ├── pku_training.utf8 └── score /README.md: -------------------------------------------------------------------------------- 1 | #Requirements: 2 | 3 | * Ubuntu 12.04 4 | * Python 2.7.3 5 | * Perl 5 6 | 7 | File Description: 8 | 9 | src/cwsFMM.py -- source code of tool cwFMM 10 | 11 | src/pku_training.utf8 -- training data 12 | 13 | src/pku_test.utf8 -- test data 14 | 15 | socre/score -- tools to socre the result 16 | 17 | socre/pku_test_gold.utf8 -- gold data 18 | 19 | ref -- reference papers of the project 20 | 21 | #Tool: 22 | 23 | 24 | ## cwsBMM 25 | 26 | Description: 27 | 28 | Using Backward Maximum Match(BMM) algorithm to do Chinese Word Segamentation 29 | 30 | === TOTAL TRUE WORDS RECALL: 0.924 31 | 32 | === TOTAL TEST WORDS PRECISION: 0.897 33 | 34 | === F MEASURE: 0.910 35 | 36 | Usage: 37 | 38 | python cwsBMM.py training_file test_file result_file 39 | 40 | perl score training_file gold_file result_file > score_file 41 | 42 | Notice: 43 | All the data and the tool score come from:http://sighan.cs.uchicago.edu/bakeoff2005/ 44 | 45 | 46 | 47 | ## cwsFMM 48 | 49 | Description: 50 | 51 | Using Forward Maximum Match(BMM) algorithm to do Chinese Word Segamentation 52 | 53 | === TOTAL TRUE WORDS RECALL: 0.920 54 | 55 | === TOTAL TEST WORDS PRECISION: 0.895 56 | 57 | === F MEASURE: 0.907 58 | 59 | Usage: 60 | 61 | python cwsFMM.py training_file test_file result_file 62 | 63 | perl score training_file gold_file result_file > score_file 64 | 65 | 66 | Notice: 67 | All the data and the tool score come from:http://sighan.cs.uchicago.edu/bakeoff2005/ 68 | -------------------------------------------------------------------------------- /ref/A Conditional Random Field Word Segmenter.pdf: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minixalpha/PyCWS/aa8e6c3302a025424bc23c294ecb2b5bca2024d0/ref/A Conditional Random Field Word Segmenter.pdf -------------------------------------------------------------------------------- /ref/A Maximum Entropy Approach to Chinese Word Segmentation.pdf: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minixalpha/PyCWS/aa8e6c3302a025424bc23c294ecb2b5bca2024d0/ref/A Maximum Entropy Approach to Chinese Word Segmentation.pdf -------------------------------------------------------------------------------- /ref/[2007,黄昌宁，赵海]中文分词十年回顾.pdf: -------------------------------------------------------------------------------- https://raw.githubusercontent.com/minixalpha/PyCWS/aa8e6c3302a025424bc23c294ecb2b5bca2024d0/ref/[2007,黄昌宁，赵海]中文分词十年回顾.pdf -------------------------------------------------------------------------------- /score/score: -------------------------------------------------------------------------------- 1 | #!/usr/bin/perl -w 2 | 3 | ########################################################################### 4 | # # 5 | # SIGHAN # 6 | # Copyright (c) 2003,2005 # 7 | # All Rights Reserved. # 8 | # # 9 | # Permission is hereby granted, free of charge, to use and distribute # 10 | # this software and its documentation without restriction, including # 11 | # without limitation the rights to use, copy, modify, merge, publish, # 12 | # distribute, sublicense, and/or sell copies of this work, and to # 13 | # permit persons to whom this work is furnished to do so, subject to # 14 | # the following conditions: # 15 | # 1. The code must retain the above copyright notice, this list of # 16 | # conditions and the following disclaimer. # 17 | # 2. Any modifications must be clearly marked as such. # 18 | # 3. Original authors' names are not deleted. # 19 | # 4. The authors' names are not used to endorse or promote products # 20 | # derived from this software without specific prior written # 21 | # permission. # 22 | # # 23 | # SIGHAN AND THE CONTRIBUTORS TO THIS WORK DISCLAIM ALL WARRANTIES # 24 | # WITH REGARD TO THIS SOFTWARE, INCLUDING ALL IMPLIED WARRANTIES OF # 25 | # MERCHANTABILITY AND FITNESS, IN NO EVENT SHALL SIGHAN NOR THE # 26 | # CONTRIBUTORS BE LIABLE FOR ANY SPECIAL, INDIRECT OR CONSEQUENTIAL # 27 | # DAMAGES OR ANY DAMAGES WHATSOEVER RESULTING FROM LOSS OF USE, DATA # 28 | # OR PROFITS, WHETHER IN AN ACTION OF CONTRACT, NEGLIGENCE OR OTHER # 29 | # TORTIOUS ACTION, ARISING OUT OF OR IN CONNECTION WITH THE USE OR # 30 | # PERFORMANCE OF THIS SOFTWARE. # 31 | # # 32 | ########################################################################### 33 | # # 34 | # Author: Richard Sproat (rws@uiuc.edu) # 35 | # Tom Emerson (tree@basistech.com) # 36 | # # 37 | ########################################################################### 38 | 39 | ## This code depends upon a version of diff (e.g. GNU diffutils 2.7.2) 40 | ## that supports the -y flag: 41 | ## 42 | ## -y Use the side by side output format. 43 | ## 44 | ## change the following per your installation: 45 | 46 | $diff = "/usr/bin/diff"; 47 | 48 | $USAGE = "Usage:\t$0 dictionary truth test\n\t"; 49 | 50 | if (@ARGV != 3) {print "$USAGE\n"; exit;} 51 | 52 | $tmp1 = "/tmp/comp01$$"; 53 | $tmp2 = "/tmp/comp02$$"; 54 | 55 | %dict = (); 56 | 57 | open (S, $ARGV[0]) or die "$ARGV[0]: $!\n"; 58 | 59 | while () { 60 | chop; 61 | s/^\s*//; 62 | s/\s*$//; 63 | $dict{$_} = 1; 64 | } 65 | 66 | close(S); 67 | 68 | open (TRUTH, $ARGV[1]) or die "$ARGV[1]: $!\n"; 69 | open (TEST, $ARGV[2]) or die "$ARGV[2]: $!\n"; 70 | 71 | $Tot = $Del = $Ins = $Subst = $Truecount = $Testcount = 0; 72 | $RawRecall = $RawPrecision = 0; 73 | 74 | $linenum = 0; 75 | 76 | 77 | $IVMISSED = $OOVMISSED = $OOV = $IV = 0; 78 | 79 | $file1 = $ARGV[1]; 80 | $file2 = $ARGV[2]; 81 | $file1 =~ s=^/.*/==; 82 | $file2 =~ s=^/.*/==; 83 | 84 | while (defined($truth = ) && defined($test = )) { 85 | $truth =~ s/^\s*//; 86 | $test =~ s/^\s*//; 87 | $truth =~ s/\s*$//; 88 | $test =~ s/\s*$//; 89 | $truth =~ s/(\xe3\x80\x80)|(\xa1\x40)/ /g; 90 | $test =~ s/(\xe3\x80\x80)|(\xa1\x40)/ /g; 91 | $truth =~ s/ //g; 92 | $test =~ s/ //g; 93 | @truthwords = split /\s+/, $truth; 94 | @testwords = split /\s+/, $test; 95 | $truecount = scalar(@truthwords); 96 | $testcount = scalar(@testwords); 97 | ++$linenum; 98 | if ($truecount == 0) { 99 | if ($testcount > 0) { 100 | print STDERR "Warning: training is 0 but test is nonzero, possible misalignment at line $linenum.\n"; 101 | } 102 | next; 103 | } 104 | if ($testcount == 0) { 105 | print STDERR "Warning: No output in test data where there is in training data, line $linenum\n"; 106 | } 107 | open (T1, ">$tmp1") or die "Can't open $tmp1"; 108 | open (T2, ">$tmp2") or die "Can't open $tmp2"; 109 | foreach my $w (@truthwords) { print T1 "$w\n"; } 110 | foreach my $w (@testwords) {print T2 "$w\n";} 111 | close (T1); 112 | close (T2); 113 | open (P, "$diff -y $tmp1 $tmp2 |") 114 | or die "Can't open pipe.\n"; 115 | print "--$file1-------$file2----$linenum\n"; 116 | my $del = 0; 117 | my $ins = 0; 118 | my $subst = 0; 119 | my $rawrecall = 0; 120 | my $rawprecision = 0; 121 | while (

) { 122 | my $err = 0; 123 | if (/\s\|\s/) {$subst++ ; $err++; } 124 | elsif (/\s\>\s/) {$ins++ ; $err++; } 125 | elsif (/\s\<\s/) {$del++ ; $err++; } 126 | if (/^([^\s]+)\s/) { 127 | my $w = $1; 128 | if (!$dict{$w}) {++$OOV;} 129 | else {++$IV;} 130 | if (/^[^\s]+\s.*\s[\|\>\<]\s/) { 131 | if (!$dict{$w}) {++$OOVMISSED;} 132 | else {++$IVMISSED;} 133 | ++$rawrecall; 134 | } 135 | } 136 | if (/\s[\|\>\<]\s.*[^\s]$/) { ++$rawprecision; } 137 | print "$_"; 138 | } 139 | close (P); 140 | my $tot = $del + $ins + $subst; 141 | $Tot += $tot; 142 | $Del += $del; 143 | $Ins += $ins; 144 | $Subst += $subst; 145 | $Truecount += $truecount; 146 | $Testcount += $testcount; 147 | $rawrecall = $truecount - $rawrecall; 148 | $rawprecision = $testcount - $rawprecision; 149 | $RawRecall += $rawrecall; 150 | $RawPrecision += $rawprecision; 151 | $rawrecall = sprintf("%2.3f", $rawrecall/$truecount); 152 | $rawprecision = sprintf("%2.3f", $rawprecision/$testcount); 153 | print "INSERTIONS:\t$ins\n"; 154 | print "DELETIONS:\t$del\n"; 155 | print "SUBSTITUTIONS:\t$subst\n"; 156 | print "NCHANGE:\t$tot\n"; 157 | print "NTRUTH:\t$truecount\n"; 158 | print "NTEST:\t$testcount\n"; 159 | print "TRUE WORDS RECALL:\t$rawrecall\n"; 160 | print "TEST WORDS PRECISION:\t$rawprecision\n"; 161 | } 162 | 163 | close(TRUTH); 164 | close(TEST); 165 | unlink($tmp1); 166 | unlink($tmp2); 167 | 168 | print "=== SUMMARY:\n"; 169 | print "=== TOTAL INSERTIONS:\t$Ins\n"; 170 | print "=== TOTAL DELETIONS:\t$Del\n"; 171 | print "=== TOTAL SUBSTITUTIONS:\t$Subst\n"; 172 | print "=== TOTAL NCHANGE:\t$Tot\n"; 173 | print "=== TOTAL TRUE WORD COUNT:\t$Truecount\n"; 174 | print "=== TOTAL TEST WORD COUNT:\t$Testcount\n"; 175 | $RawRecall = $RawRecall/$Truecount; 176 | $RawPrecision = $RawPrecision/$Testcount; 177 | $beta = 1; 178 | $R = $RawRecall; 179 | $P = $RawPrecision; 180 | $F = (1 + $beta)*$P*$R/($beta * $P + $R); 181 | $F = sprintf("%2.3f", $F); 182 | $RawRecall = sprintf("%2.3f", $RawRecall); 183 | $RawPrecision = sprintf("%2.3f", $RawPrecision); 184 | print "=== TOTAL TRUE WORDS RECALL:\t$RawRecall\n"; 185 | print "=== TOTAL TEST WORDS PRECISION:\t$RawPrecision\n"; 186 | print "=== F MEASURE:\t$F\n"; 187 | if ($OOV > 0) { 188 | $OOVMISSED = sprintf("%2.3f", 1 - $OOVMISSED / $OOV); 189 | } 190 | else { 191 | $OOVMISSED = "--"; 192 | } 193 | $OOV = sprintf("%2.3f", $OOV / $Truecount); 194 | if ($IV > 0) { 195 | $IVMISSED = sprintf("%2.3f", 1 - $IVMISSED / $IV); 196 | } 197 | else { 198 | $IVMISSED = "--"; 199 | } 200 | print "=== OOV Rate:\t$OOV\n"; 201 | print "=== OOV Recall Rate:\t$OOVMISSED\n"; 202 | print "=== IV Recall Rate:\t$IVMISSED\n"; 203 | 204 | print "###\t$file2\t$Ins\t$Del\t$Subst\t$Tot\t$Truecount\t$Testcount\t$RawRecall\t$RawPrecision\t$F\t$OOV\t$OOVMISSED\t$IVMISSED\n"; 205 | exit(0); 206 | -------------------------------------------------------------------------------- /src/LICENSE: -------------------------------------------------------------------------------- 1 | GNU GENERAL PUBLIC LICENSE 2 | Version 3, 29 June 2007 3 | 4 | Copyright (C) 2007 Free Software Foundation, Inc. 5 | Everyone is permitted to copy and distribute verbatim copies 6 | of this license document, but changing it is not allowed. 7 | 8 | Preamble 9 | 10 | The GNU General Public License is a free, copyleft license for 11 | software and other kinds of works. 12 | 13 | The licenses for most software and other practical works are designed 14 | to take away your freedom to share and change the works. By contrast, 15 | the GNU General Public License is intended to guarantee your freedom to 16 | share and change all versions of a program--to make sure it remains free 17 | software for all its users. We, the Free Software Foundation, use the 18 | GNU General Public License for most of our software; it applies also to 19 | any other work released this way by its authors. You can apply it to 20 | your programs, too. 21 | 22 | When we speak of free software, we are referring to freedom, not 23 | price. Our General Public Licenses are designed to make sure that you 24 | have the freedom to distribute copies of free software (and charge for 25 | them if you wish), that you receive source code or can get it if you 26 | want it, that you can change the software or use pieces of it in new 27 | free programs, and that you know you can do these things. 28 | 29 | To protect your rights, we need to prevent others from denying you 30 | these rights or asking you to surrender the rights. Therefore, you have 31 | certain responsibilities if you distribute copies of the software, or if 32 | you modify it: responsibilities to respect the freedom of others. 33 | 34 | For example, if you distribute copies of such a program, whether 35 | gratis or for a fee, you must pass on to the recipients the same 36 | freedoms that you received. You must make sure that they, too, receive 37 | or can get the source code. And you must show them these terms so they 38 | know their rights. 39 | 40 | Developers that use the GNU GPL protect your rights with two steps: 41 | (1) assert copyright on the software, and (2) offer you this License 42 | giving you legal permission to copy, distribute and/or modify it. 43 | 44 | For the developers' and authors' protection, the GPL clearly explains 45 | that there is no warranty for this free software. For both users' and 46 | authors' sake, the GPL requires that modified versions be marked as 47 | changed, so that their problems will not be attributed erroneously to 48 | authors of previous versions. 49 | 50 | Some devices are designed to deny users access to install or run 51 | modified versions of the software inside them, although the manufacturer 52 | can do so. This is fundamentally incompatible with the aim of 53 | protecting users' freedom to change the software. The systematic 54 | pattern of such abuse occurs in the area of products for individuals to 55 | use, which is precisely where it is most unacceptable. Therefore, we 56 | have designed this version of the GPL to prohibit the practice for those 57 | products. If such problems arise substantially in other domains, we 58 | stand ready to extend this provision to those domains in future versions 59 | of the GPL, as needed to protect the freedom of users. 60 | 61 | Finally, every program is threatened constantly by software patents. 62 | States should not allow patents to restrict development and use of 63 | software on general-purpose computers, but in those that do, we wish to 64 | avoid the special danger that patents applied to a free program could 65 | make it effectively proprietary. To prevent this, the GPL assures that 66 | patents cannot be used to render the program non-free. 67 | 68 | The precise terms and conditions for copying, distribution and 69 | modification follow. 70 | 71 | TERMS AND CONDITIONS 72 | 73 | 0. Definitions. 74 | 75 | "This License" refers to version 3 of the GNU General Public License. 76 | 77 | "Copyright" also means copyright-like laws that apply to other kinds of 78 | works, such as semiconductor masks. 79 | 80 | "The Program" refers to any copyrightable work licensed under this 81 | License. Each licensee is addressed as "you". "Licensees" and 82 | "recipients" may be individuals or organizations. 83 | 84 | To "modify" a work means to copy from or adapt all or part of the work 85 | in a fashion requiring copyright permission, other than the making of an 86 | exact copy. The resulting work is called a "modified version" of the 87 | earlier work or a work "based on" the earlier work. 88 | 89 | A "covered work" means either the unmodified Program or a work based 90 | on the Program. 91 | 92 | To "propagate" a work means to do anything with it that, without 93 | permission, would make you directly or secondarily liable for 94 | infringement under applicable copyright law, except executing it on a 95 | computer or modifying a private copy. Propagation includes copying, 96 | distribution (with or without modification), making available to the 97 | public, and in some countries other activities as well. 98 | 99 | To "convey" a work means any kind of propagation that enables other 100 | parties to make or receive copies. Mere interaction with a user through 101 | a computer network, with no transfer of a copy, is not conveying. 102 | 103 | An interactive user interface displays "Appropriate Legal Notices" 104 | to the extent that it includes a convenient and prominently visible 105 | feature that (1) displays an appropriate copyright notice, and (2) 106 | tells the user that there is no warranty for the work (except to the 107 | extent that warranties are provided), that licensees may convey the 108 | work under this License, and how to view a copy of this License. If 109 | the interface presents a list of user commands or options, such as a 110 | menu, a prominent item in the list meets this criterion. 111 | 112 | 1. Source Code. 113 | 114 | The "source code" for a work means the preferred form of the work 115 | for making modifications to it. "Object code" means any non-source 116 | form of a work. 117 | 118 | A "Standard Interface" means an interface that either is an official 119 | standard defined by a recognized standards body, or, in the case of 120 | interfaces specified for a particular programming language, one that 121 | is widely used among developers working in that language. 122 | 123 | The "System Libraries" of an executable work include anything, other 124 | than the work as a whole, that (a) is included in the normal form of 125 | packaging a Major Component, but which is not part of that Major 126 | Component, and (b) serves only to enable use of the work with that 127 | Major Component, or to implement a Standard Interface for which an 128 | implementation is available to the public in source code form. A 129 | "Major Component", in this context, means a major essential component 130 | (kernel, window system, and so on) of the specific operating system 131 | (if any) on which the executable work runs, or a compiler used to 132 | produce the work, or an object code interpreter used to run it. 133 | 134 | The "Corresponding Source" for a work in object code form means all 135 | the source code needed to generate, install, and (for an executable 136 | work) run the object code and to modify the work, including scripts to 137 | control those activities. However, it does not include the work's 138 | System Libraries, or general-purpose tools or generally available free 139 | programs which are used unmodified in performing those activities but 140 | which are not part of the work. For example, Corresponding Source 141 | includes interface definition files associated with source files for 142 | the work, and the source code for shared libraries and dynamically 143 | linked subprograms that the work is specifically designed to require, 144 | such as by intimate data communication or control flow between those 145 | subprograms and other parts of the work. 146 | 147 | The Corresponding Source need not include anything that users 148 | can regenerate automatically from other parts of the Corresponding 149 | Source. 150 | 151 | The Corresponding Source for a work in source code form is that 152 | same work. 153 | 154 | 2. Basic Permissions. 155 | 156 | All rights granted under this License are granted for the term of 157 | copyright on the Program, and are irrevocable provided the stated 158 | conditions are met. This License explicitly affirms your unlimited 159 | permission to run the unmodified Program. The output from running a 160 | covered work is covered by this License only if the output, given its 161 | content, constitutes a covered work. This License acknowledges your 162 | rights of fair use or other equivalent, as provided by copyright law. 163 | 164 | You may make, run and propagate covered works that you do not 165 | convey, without conditions so long as your license otherwise remains 166 | in force. You may convey covered works to others for the sole purpose 167 | of having them make modifications exclusively for you, or provide you 168 | with facilities for running those works, provided that you comply with 169 | the terms of this License in conveying all material for which you do 170 | not control copyright. Those thus making or running the covered works 171 | for you must do so exclusively on your behalf, under your direction 172 | and control, on terms that prohibit them from making any copies of 173 | your copyrighted material outside their relationship with you. 174 | 175 | Conveying under any other circumstances is permitted solely under 176 | the conditions stated below. Sublicensing is not allowed; section 10 177 | makes it unnecessary. 178 | 179 | 3. Protecting Users' Legal Rights From Anti-Circumvention Law. 180 | 181 | No covered work shall be deemed part of an effective technological 182 | measure under any applicable law fulfilling obligations under article 183 | 11 of the WIPO copyright treaty adopted on 20 December 1996, or 184 | similar laws prohibiting or restricting circumvention of such 185 | measures. 186 | 187 | When you convey a covered work, you waive any legal power to forbid 188 | circumvention of technological measures to the extent such circumvention 189 | is effected by exercising rights under this License with respect to 190 | the covered work, and you disclaim any intention to limit operation or 191 | modification of the work as a means of enforcing, against the work's 192 | users, your or third parties' legal rights to forbid circumvention of 193 | technological measures. 194 | 195 | 4. Conveying Verbatim Copies. 196 | 197 | You may convey verbatim copies of the Program's source code as you 198 | receive it, in any medium, provided that you conspicuously and 199 | appropriately publish on each copy an appropriate copyright notice; 200 | keep intact all notices stating that this License and any 201 | non-permissive terms added in accord with section 7 apply to the code; 202 | keep intact all notices of the absence of any warranty; and give all 203 | recipients a copy of this License along with the Program. 204 | 205 | You may charge any price or no price for each copy that you convey, 206 | and you may offer support or warranty protection for a fee. 207 | 208 | 5. Conveying Modified Source Versions. 209 | 210 | You may convey a work based on the Program, or the modifications to 211 | produce it from the Program, in the form of source code under the 212 | terms of section 4, provided that you also meet all of these conditions: 213 | 214 | a) The work must carry prominent notices stating that you modified 215 | it, and giving a relevant date. 216 | 217 | b) The work must carry prominent notices stating that it is 218 | released under this License and any conditions added under section 219 | 7. This requirement modifies the requirement in section 4 to 220 | "keep intact all notices". 221 | 222 | c) You must license the entire work, as a whole, under this 223 | License to anyone who comes into possession of a copy. This 224 | License will therefore apply, along with any applicable section 7 225 | additional terms, to the whole of the work, and all its parts, 226 | regardless of how they are packaged. This License gives no 227 | permission to license the work in any other way, but it does not 228 | invalidate such permission if you have separately received it. 229 | 230 | d) If the work has interactive user interfaces, each must display 231 | Appropriate Legal Notices; however, if the Program has interactive 232 | interfaces that do not display Appropriate Legal Notices, your 233 | work need not make them do so. 234 | 235 | A compilation of a covered work with other separate and independent 236 | works, which are not by their nature extensions of the covered work, 237 | and which are not combined with it such as to form a larger program, 238 | in or on a volume of a storage or distribution medium, is called an 239 | "aggregate" if the compilation and its resulting copyright are not 240 | used to limit the access or legal rights of the compilation's users 241 | beyond what the individual works permit. Inclusion of a covered work 242 | in an aggregate does not cause this License to apply to the other 243 | parts of the aggregate. 244 | 245 | 6. Conveying Non-Source Forms. 246 | 247 | You may convey a covered work in object code form under the terms 248 | of sections 4 and 5, provided that you also convey the 249 | machine-readable Corresponding Source under the terms of this License, 250 | in one of these ways: 251 | 252 | a) Convey the object code in, or embodied in, a physical product 253 | (including a physical distribution medium), accompanied by the 254 | Corresponding Source fixed on a durable physical medium 255 | customarily used for software interchange. 256 | 257 | b) Convey the object code in, or embodied in, a physical product 258 | (including a physical distribution medium), accompanied by a 259 | written offer, valid for at least three years and valid for as 260 | long as you offer spare parts or customer support for that product 261 | model, to give anyone who possesses the object code either (1) a 262 | copy of the Corresponding Source for all the software in the 263 | product that is covered by this License, on a durable physical 264 | medium customarily used for software interchange, for a price no 265 | more than your reasonable cost of physically performing this 266 | conveying of source, or (2) access to copy the 267 | Corresponding Source from a network server at no charge. 268 | 269 | c) Convey individual copies of the object code with a copy of the 270 | written offer to provide the Corresponding Source. This 271 | alternative is allowed only occasionally and noncommercially, and 272 | only if you received the object code with such an offer, in accord 273 | with subsection 6b. 274 | 275 | d) Convey the object code by offering access from a designated 276 | place (gratis or for a charge), and offer equivalent access to the 277 | Corresponding Source in the same way through the same place at no 278 | further charge. You need not require recipients to copy the 279 | Corresponding Source along with the object code. If the place to 280 | copy the object code is a network server, the Corresponding Source 281 | may be on a different server (operated by you or a third party) 282 | that supports equivalent copying facilities, provided you maintain 283 | clear directions next to the object code saying where to find the 284 | Corresponding Source. Regardless of what server hosts the 285 | Corresponding Source, you remain obligated to ensure that it is 286 | available for as long as needed to satisfy these requirements. 287 | 288 | e) Convey the object code using peer-to-peer transmission, provided 289 | you inform other peers where the object code and Corresponding 290 | Source of the work are being offered to the general public at no 291 | charge under subsection 6d. 292 | 293 | A separable portion of the object code, whose source code is excluded 294 | from the Corresponding Source as a System Library, need not be 295 | included in conveying the object code work. 296 | 297 | A "User Product" is either (1) a "consumer product", which means any 298 | tangible personal property which is normally used for personal, family, 299 | or household purposes, or (2) anything designed or sold for incorporation 300 | into a dwelling. In determining whether a product is a consumer product, 301 | doubtful cases shall be resolved in favor of coverage. For a particular 302 | product received by a particular user, "normally used" refers to a 303 | typical or common use of that class of product, regardless of the status 304 | of the particular user or of the way in which the particular user 305 | actually uses, or expects or is expected to use, the product. A product 306 | is a consumer product regardless of whether the product has substantial 307 | commercial, industrial or non-consumer uses, unless such uses represent 308 | the only significant mode of use of the product. 309 | 310 | "Installation Information" for a User Product means any methods, 311 | procedures, authorization keys, or other information required to install 312 | and execute modified versions of a covered work in that User Product from 313 | a modified version of its Corresponding Source. The information must 314 | suffice to ensure that the continued functioning of the modified object 315 | code is in no case prevented or interfered with solely because 316 | modification has been made. 317 | 318 | If you convey an object code work under this section in, or with, or 319 | specifically for use in, a User Product, and the conveying occurs as 320 | part of a transaction in which the right of possession and use of the 321 | User Product is transferred to the recipient in perpetuity or for a 322 | fixed term (regardless of how the transaction is characterized), the 323 | Corresponding Source conveyed under this section must be accompanied 324 | by the Installation Information. But this requirement does not apply 325 | if neither you nor any third party retains the ability to install 326 | modified object code on the User Product (for example, the work has 327 | been installed in ROM). 328 | 329 | The requirement to provide Installation Information does not include a 330 | requirement to continue to provide support service, warranty, or updates 331 | for a work that has been modified or installed by the recipient, or for 332 | the User Product in which it has been modified or installed. Access to a 333 | network may be denied when the modification itself materially and 334 | adversely affects the operation of the network or violates the rules and 335 | protocols for communication across the network. 336 | 337 | Corresponding Source conveyed, and Installation Information provided, 338 | in accord with this section must be in a format that is publicly 339 | documented (and with an implementation available to the public in 340 | source code form), and must require no special password or key for 341 | unpacking, reading or copying. 342 | 343 | 7. Additional Terms. 344 | 345 | "Additional permissions" are terms that supplement the terms of this 346 | License by making exceptions from one or more of its conditions. 347 | Additional permissions that are applicable to the entire Program shall 348 | be treated as though they were included in this License, to the extent 349 | that they are valid under applicable law. If additional permissions 350 | apply only to part of the Program, that part may be used separately 351 | under those permissions, but the entire Program remains governed by 352 | this License without regard to the additional permissions. 353 | 354 | When you convey a copy of a covered work, you may at your option 355 | remove any additional permissions from that copy, or from any part of 356 | it. (Additional permissions may be written to require their own 357 | removal in certain cases when you modify the work.) You may place 358 | additional permissions on material, added by you to a covered work, 359 | for which you have or can give appropriate copyright permission. 360 | 361 | Notwithstanding any other provision of this License, for material you 362 | add to a covered work, you may (if authorized by the copyright holders of 363 | that material) supplement the terms of this License with terms: 364 | 365 | a) Disclaiming warranty or limiting liability differently from the 366 | terms of sections 15 and 16 of this License; or 367 | 368 | b) Requiring preservation of specified reasonable legal notices or 369 | author attributions in that material or in the Appropriate Legal 370 | Notices displayed by works containing it; or 371 | 372 | c) Prohibiting misrepresentation of the origin of that material, or 373 | requiring that modified versions of such material be marked in 374 | reasonable ways as different from the original version; or 375 | 376 | d) Limiting the use for publicity purposes of names of licensors or 377 | authors of the material; or 378 | 379 | e) Declining to grant rights under trademark law for use of some 380 | trade names, trademarks, or service marks; or 381 | 382 | f) Requiring indemnification of licensors and authors of that 383 | material by anyone who conveys the material (or modified versions of 384 | it) with contractual assumptions of liability to the recipient, for 385 | any liability that these contractual assumptions directly impose on 386 | those licensors and authors. 387 | 388 | All other non-permissive additional terms are considered "further 389 | restrictions" within the meaning of section 10. If the Program as you 390 | received it, or any part of it, contains a notice stating that it is 391 | governed by this License along with a term that is a further 392 | restriction, you may remove that term. If a license document contains 393 | a further restriction but permits relicensing or conveying under this 394 | License, you may add to a covered work material governed by the terms 395 | of that license document, provided that the further restriction does 396 | not survive such relicensing or conveying. 397 | 398 | If you add terms to a covered work in accord with this section, you 399 | must place, in the relevant source files, a statement of the 400 | additional terms that apply to those files, or a notice indicating 401 | where to find the applicable terms. 402 | 403 | Additional terms, permissive or non-permissive, may be stated in the 404 | form of a separately written license, or stated as exceptions; 405 | the above requirements apply either way. 406 | 407 | 8. Termination. 408 | 409 | You may not propagate or modify a covered work except as expressly 410 | provided under this License. Any attempt otherwise to propagate or 411 | modify it is void, and will automatically terminate your rights under 412 | this License (including any patent licenses granted under the third 413 | paragraph of section 11). 414 | 415 | However, if you cease all violation of this License, then your 416 | license from a particular copyright holder is reinstated (a) 417 | provisionally, unless and until the copyright holder explicitly and 418 | finally terminates your license, and (b) permanently, if the copyright 419 | holder fails to notify you of the violation by some reasonable means 420 | prior to 60 days after the cessation. 421 | 422 | Moreover, your license from a particular copyright holder is 423 | reinstated permanently if the copyright holder notifies you of the 424 | violation by some reasonable means, this is the first time you have 425 | received notice of violation of this License (for any work) from that 426 | copyright holder, and you cure the violation prior to 30 days after 427 | your receipt of the notice. 428 | 429 | Termination of your rights under this section does not terminate the 430 | licenses of parties who have received copies or rights from you under 431 | this License. If your rights have been terminated and not permanently 432 | reinstated, you do not qualify to receive new licenses for the same 433 | material under section 10. 434 | 435 | 9. Acceptance Not Required for Having Copies. 436 | 437 | You are not required to accept this License in order to receive or 438 | run a copy of the Program. Ancillary propagation of a covered work 439 | occurring solely as a consequence of using peer-to-peer transmission 440 | to receive a copy likewise does not require acceptance. However, 441 | nothing other than this License grants you permission to propagate or 442 | modify any covered work. These actions infringe copyright if you do 443 | not accept this License. Therefore, by modifying or propagating a 444 | covered work, you indicate your acceptance of this License to do so. 445 | 446 | 10. Automatic Licensing of Downstream Recipients. 447 | 448 | Each time you convey a covered work, the recipient automatically 449 | receives a license from the original licensors, to run, modify and 450 | propagate that work, subject to this License. You are not responsible 451 | for enforcing compliance by third parties with this License. 452 | 453 | An "entity transaction" is a transaction transferring control of an 454 | organization, or substantially all assets of one, or subdividing an 455 | organization, or merging organizations. If propagation of a covered 456 | work results from an entity transaction, each party to that 457 | transaction who receives a copy of the work also receives whatever 458 | licenses to the work the party's predecessor in interest had or could 459 | give under the previous paragraph, plus a right to possession of the 460 | Corresponding Source of the work from the predecessor in interest, if 461 | the predecessor has it or can get it with reasonable efforts. 462 | 463 | You may not impose any further restrictions on the exercise of the 464 | rights granted or affirmed under this License. For example, you may 465 | not impose a license fee, royalty, or other charge for exercise of 466 | rights granted under this License, and you may not initiate litigation 467 | (including a cross-claim or counterclaim in a lawsuit) alleging that 468 | any patent claim is infringed by making, using, selling, offering for 469 | sale, or importing the Program or any portion of it. 470 | 471 | 11. Patents. 472 | 473 | A "contributor" is a copyright holder who authorizes use under this 474 | License of the Program or a work on which the Program is based. The 475 | work thus licensed is called the contributor's "contributor version". 476 | 477 | A contributor's "essential patent claims" are all patent claims 478 | owned or controlled by the contributor, whether already acquired or 479 | hereafter acquired, that would be infringed by some manner, permitted 480 | by this License, of making, using, or selling its contributor version, 481 | but do not include claims that would be infringed only as a 482 | consequence of further modification of the contributor version. For 483 | purposes of this definition, "control" includes the right to grant 484 | patent sublicenses in a manner consistent with the requirements of 485 | this License. 486 | 487 | Each contributor grants you a non-exclusive, worldwide, royalty-free 488 | patent license under the contributor's essential patent claims, to 489 | make, use, sell, offer for sale, import and otherwise run, modify and 490 | propagate the contents of its contributor version. 491 | 492 | In the following three paragraphs, a "patent license" is any express 493 | agreement or commitment, however denominated, not to enforce a patent 494 | (such as an express permission to practice a patent or covenant not to 495 | sue for patent infringement). To "grant" such a patent license to a 496 | party means to make such an agreement or commitment not to enforce a 497 | patent against the party. 498 | 499 | If you convey a covered work, knowingly relying on a patent license, 500 | and the Corresponding Source of the work is not available for anyone 501 | to copy, free of charge and under the terms of this License, through a 502 | publicly available network server or other readily accessible means, 503 | then you must either (1) cause the Corresponding Source to be so 504 | available, or (2) arrange to deprive yourself of the benefit of the 505 | patent license for this particular work, or (3) arrange, in a manner 506 | consistent with the requirements of this License, to extend the patent 507 | license to downstream recipients. "Knowingly relying" means you have 508 | actual knowledge that, but for the patent license, your conveying the 509 | covered work in a country, or your recipient's use of the covered work 510 | in a country, would infringe one or more identifiable patents in that 511 | country that you have reason to believe are valid. 512 | 513 | If, pursuant to or in connection with a single transaction or 514 | arrangement, you convey, or propagate by procuring conveyance of, a 515 | covered work, and grant a patent license to some of the parties 516 | receiving the covered work authorizing them to use, propagate, modify 517 | or convey a specific copy of the covered work, then the patent license 518 | you grant is automatically extended to all recipients of the covered 519 | work and works based on it. 520 | 521 | A patent license is "discriminatory" if it does not include within 522 | the scope of its coverage, prohibits the exercise of, or is 523 | conditioned on the non-exercise of one or more of the rights that are 524 | specifically granted under this License. You may not convey a covered 525 | work if you are a party to an arrangement with a third party that is 526 | in the business of distributing software, under which you make payment 527 | to the third party based on the extent of your activity of conveying 528 | the work, and under which the third party grants, to any of the 529 | parties who would receive the covered work from you, a discriminatory 530 | patent license (a) in connection with copies of the covered work 531 | conveyed by you (or copies made from those copies), or (b) primarily 532 | for and in connection with specific products or compilations that 533 | contain the covered work, unless you entered into that arrangement, 534 | or that patent license was granted, prior to 28 March 2007. 535 | 536 | Nothing in this License shall be construed as excluding or limiting 537 | any implied license or other defenses to infringement that may 538 | otherwise be available to you under applicable patent law. 539 | 540 | 12. No Surrender of Others' Freedom. 541 | 542 | If conditions are imposed on you (whether by court order, agreement or 543 | otherwise) that contradict the conditions of this License, they do not 544 | excuse you from the conditions of this License. If you cannot convey a 545 | covered work so as to satisfy simultaneously your obligations under this 546 | License and any other pertinent obligations, then as a consequence you may 547 | not convey it at all. For example, if you agree to terms that obligate you 548 | to collect a royalty for further conveying from those to whom you convey 549 | the Program, the only way you could satisfy both those terms and this 550 | License would be to refrain entirely from conveying the Program. 551 | 552 | 13. Use with the GNU Affero General Public License. 553 | 554 | Notwithstanding any other provision of this License, you have 555 | permission to link or combine any covered work with a work licensed 556 | under version 3 of the GNU Affero General Public License into a single 557 | combined work, and to convey the resulting work. The terms of this 558 | License will continue to apply to the part which is the covered work, 559 | but the special requirements of the GNU Affero General Public License, 560 | section 13, concerning interaction through a network will apply to the 561 | combination as such. 562 | 563 | 14. Revised Versions of this License. 564 | 565 | The Free Software Foundation may publish revised and/or new versions of 566 | the GNU General Public License from time to time. Such new versions will 567 | be similar in spirit to the present version, but may differ in detail to 568 | address new problems or concerns. 569 | 570 | Each version is given a distinguishing version number. If the 571 | Program specifies that a certain numbered version of the GNU General 572 | Public License "or any later version" applies to it, you have the 573 | option of following the terms and conditions either of that numbered 574 | version or of any later version published by the Free Software 575 | Foundation. If the Program does not specify a version number of the 576 | GNU General Public License, you may choose any version ever published 577 | by the Free Software Foundation. 578 | 579 | If the Program specifies that a proxy can decide which future 580 | versions of the GNU General Public License can be used, that proxy's 581 | public statement of acceptance of a version permanently authorizes you 582 | to choose that version for the Program. 583 | 584 | Later license versions may give you additional or different 585 | permissions. However, no additional obligations are imposed on any 586 | author or copyright holder as a result of your choosing to follow a 587 | later version. 588 | 589 | 15. Disclaimer of Warranty. 590 | 591 | THERE IS NO WARRANTY FOR THE PROGRAM, TO THE EXTENT PERMITTED BY 592 | APPLICABLE LAW. EXCEPT WHEN OTHERWISE STATED IN WRITING THE COPYRIGHT 593 | HOLDERS AND/OR OTHER PARTIES PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY 594 | OF ANY KIND, EITHER EXPRESSED OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, 595 | THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR 596 | PURPOSE. THE ENTIRE RISK AS TO THE QUALITY AND PERFORMANCE OF THE PROGRAM 597 | IS WITH YOU. SHOULD THE PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF 598 | ALL NECESSARY SERVICING, REPAIR OR CORRECTION. 599 | 600 | 16. Limitation of Liability. 601 | 602 | IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING 603 | WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MODIFIES AND/OR CONVEYS 604 | THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES, INCLUDING ANY 605 | GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING OUT OF THE 606 | USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED TO LOSS OF 607 | DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY YOU OR THIRD 608 | PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER PROGRAMS), 609 | EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE POSSIBILITY OF 610 | SUCH DAMAGES. 611 | 612 | 17. Interpretation of Sections 15 and 16. 613 | 614 | If the disclaimer of warranty and limitation of liability provided 615 | above cannot be given local legal effect according to their terms, 616 | reviewing courts shall apply local law that most closely approximates 617 | an absolute waiver of all civil liability in connection with the 618 | Program, unless a warranty or assumption of liability accompanies a 619 | copy of the Program in return for a fee. 620 | 621 | END OF TERMS AND CONDITIONS 622 | 623 | How to Apply These Terms to Your New Programs 624 | 625 | If you develop a new program, and you want it to be of the greatest 626 | possible use to the public, the best way to achieve this is to make it 627 | free software which everyone can redistribute and change under these terms. 628 | 629 | To do so, attach the following notices to the program. It is safest 630 | to attach them to the start of each source file to most effectively 631 | state the exclusion of warranty; and each file should have at least 632 | the "copyright" line and a pointer to where the full notice is found. 633 | 634 | 635 | Copyright (C) 636 | 637 | This program is free software: you can redistribute it and/or modify 638 | it under the terms of the GNU General Public License as published by 639 | the Free Software Foundation, either version 3 of the License, or 640 | (at your option) any later version. 641 | 642 | This program is distributed in the hope that it will be useful, 643 | but WITHOUT ANY WARRANTY; without even the implied warranty of 644 | MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the 645 | GNU General Public License for more details. 646 | 647 | You should have received a copy of the GNU General Public License 648 | along with this program. If not, see . 649 | 650 | Also add information on how to contact you by electronic and paper mail. 651 | 652 | If the program does terminal interaction, make it output a short 653 | notice like this when it starts in an interactive mode: 654 | 655 | Copyright (C) 656 | This program comes with ABSOLUTELY NO WARRANTY; for details type `show w'. 657 | This is free software, and you are welcome to redistribute it 658 | under certain conditions; type `show c' for details. 659 | 660 | The hypothetical commands `show w' and `show c' should show the appropriate 661 | parts of the General Public License. Of course, your program's commands 662 | might be different; for a GUI interface, you would use an "about box". 663 | 664 | You should also get your employer (if you work as a programmer) or school, 665 | if any, to sign a "copyright disclaimer" for the program, if necessary. 666 | For more information on this, and how to apply and follow the GNU GPL, see 667 | . 668 | 669 | The GNU General Public License does not permit incorporating your program 670 | into proprietary programs. If your program is a subroutine library, you 671 | may consider it more useful to permit linking proprietary applications with 672 | the library. If this is what you want to do, use the GNU Lesser General 673 | Public License instead of this License. But first, please read 674 | . 675 | -------------------------------------------------------------------------------- /src/ReadMe: -------------------------------------------------------------------------------- 1 | 1 系统环境： 2 | Ubuntu 12.04 3 | Python 2.7 4 | Perl 5 5 | Maxent 6 | 7 | 2 文件说明： 8 | cwsFMM_NRule.py -- 不含规则的正向最大匹配算法 9 | cwsFMM.py -- 包含规则的正向最大匹配算法 10 | cwsBMM_NRule.py -- 不含规则的逆向最大匹配算法 11 | cwsBMM.py -- 包含规则的逆向最大匹配算法 12 | cwsMaxEn-13f.py -- 13特征最大熵字标注算法 13 | cwsMaxEn-16f.py -- 16特征最大熵字标注算法 14 | pku_test.utf8 -- 测试文件 15 | pku_training.utf8 -- 训练文件 16 | pku_test_gold.utf8 -- 测试文件标准分词结果 17 | score -- 评分工具 18 | 19 | 3 工具使用 20 | 分词： 21 | Usage: python toolName training_file test_file result_file 22 | eg: 23 | python cwsFMM.py pku_training.utf8 pku_test.utf8 pku_test.result 24 | 评分： 25 | perl score training_file pku_test_gold.utf8 result_file 26 | 27 | 4 版权声明 28 | pku_test.utf8, pku_training.utf8, score 均来自Bakeoff-2005官方网站：http://www.sighan.org/bakeoff2005/　版权归Bakeoff-2005所有 29 | 其余文件遵循GPL协议, 见LICENSE文件 30 | 31 | 5 感谢 32 | 工具的开发使用了张乐博士的Maxent工具包, 在此特别感谢 33 | 工具主页：http://homepages.inf.ed.ac.uk/lzhang10/maxent_toolkit.html 34 | 工具Github主页:https://github.com/lzhang10/maxent 35 | 36 | -------------------------------------------------------------------------------- /src/autoScore.py: -------------------------------------------------------------------------------- 1 | #! /usr/bin/env python 2 | # -*- coding: utf-8 -*- 3 | 4 | # Author: minix 5 | # Date: 2013-05-18 6 | # Email: minix007@foxmail.com 7 | 8 | # 功能： 9 | # 0. 自动为多个文件评分 10 | # 1. 提取多个评分结果 11 | # 2. 打印所有评分结果 12 | 13 | # 使用： 14 | # 例如分词结果文件为：pku_test.result.10,pku_test.result.50,pku_test.result.xxx 15 | # 这样文件名的模板为pku_test.result 16 | # python autoScore pku_test.result 即可将所有此类文件评分，输出评分结果 17 | 18 | import sys 19 | import os 20 | 21 | def score_files(file_name_temp): 22 | filename_list = os.popen('ls ' + file_name_temp + '.*').readlines() 23 | score_path = '/home/zhaoxk/Project/PyCWS/score/score' 24 | for filename in filename_list: 25 | filename = filename.rstrip() 26 | fileid = filename.split('.')[-1] 27 | name = filename.split('.')[0] 28 | cmd = 'perl ' + score_path + ' pku_training.utf8 pku_test_gold.utf8 ' + filename + ' > ' + name +'.score.' + fileid 29 | print cmd 30 | os.system(cmd) 31 | 32 | def extract_score_result(file_name_temp): 33 | filename_list = os.popen('ls ' + file_name_temp + '.*').readlines() 34 | fileid_list = [] 35 | for filename in filename_list: 36 | filename = filename.rstrip() 37 | fileid = filename.split('.')[-1] 38 | fileid_list.append(int(fileid)) 39 | fileid_list.sort() 40 | 41 | result_list = [] 42 | for fileid in fileid_list: 43 | filename = file_name_temp + '.' + str(fileid) 44 | cmd = 'tail -16 ' + filename 45 | result = os.popen(cmd).readlines() 46 | cur_result_list = [] 47 | for item in result[:-1]: 48 | tmp = item.rstrip().split('\t') 49 | if len(tmp) > 1: 50 | cur_result_list.append(tmp[1]) 51 | 52 | result_list.append(cur_result_list) 53 | 54 | return (fileid_list, result_list) 55 | 56 | def main(): 57 | args = sys.argv[1:] 58 | 59 | if len(args)<1: 60 | print 'Usage: python autoScore.py file_to_score' 61 | exit(-1) 62 | 63 | file_to_score = args[0] 64 | score_files(file_to_score) 65 | (id_list, re_list) = extract_score_result(file_to_score.split('.')[0] + '.score') 66 | id_len = len(id_list) 67 | i = 0 68 | print 'id' 69 | label_list = [ 70 | 'TRUE WORDS RECALL', 'TEST WORDS PRECISION', 71 | 'TOTAL INSERTIONS', 'TOTAL DELETIONS', 72 | 'TOTAL SUBSTITUTIONS', 'TOTAL NCHANGE', 73 | 'TOTAL TRUE WORD COUNT', 'TOTAL TEST WORD COUNT', 74 | 'TOTAL TRUE WORDS RECALL', 'TOTAL TEST WORDS PRECISION', 75 | 'F MEASURE', 76 | 'OOV Rate', 'OOV Recall Rate', 77 | 'IV Recall Rate' 78 | ] 79 | print label_list 80 | while i< id_len: 81 | print id_list[i] 82 | print re_list[i] 83 | i += 1 84 | 85 | if __name__ == "__main__": 86 | main() 87 | -------------------------------------------------------------------------------- /src/cwsBMM.py: -------------------------------------------------------------------------------- 1 | #! /usr/bin/env python 2 | # -*- coding: utf-8 -*- 3 | 4 | # Author: 赵晓凯 5 | # Date: 2013-06-05 6 | # Email: minix007@foxmail.com 7 | 8 | # 使用BMM 进行中文分词 9 | # 添加了规则，用于匹配数字带来的新词 10 | 11 | import codecs 12 | import sys 13 | 14 | # 由规则处理的一些特殊符号 15 | numMath = [u'0', u'1', u'2', u'3', u'4', u'5', u'6', u'7', u'8', u'9'] 16 | numMath_suffix = [u'.', u'%', u'亿', u'万', u'千', u'百', u'十', u'个'] 17 | numCn = [u'一', u'二', u'三', u'四', u'五', u'六', u'七', u'八', u'九', u'○', u'零'] 18 | numCn_suffix_date = [u'年', u'月', u'日'] 19 | numCn_suffix_unit = [u'亿', u'万', u'千', u'百', u'十', u'个'] 20 | special_char = [u'(', u')'] 21 | num_char = numMath + numCn 22 | num_suffix = numMath_suffix + numCn_suffix_unit + numCn_suffix_date 23 | 24 | def proc_num_math(line, start): 25 | """ 处理句子中出现的数学符号 """ 26 | oldstart = start 27 | start = start + 1 28 | while line[start] in numMath or line[start] in numMath_suffix: 29 | start = start + 1 30 | return start - oldstart 31 | 32 | def proc_num_cn(line, start): 33 | """ 处理句子中出现的中文数字 """ 34 | oldstart = start 35 | while line[start] in numCn or line[start] in numCn_suffix_unit: 36 | start = start + 1 37 | return start - oldstart 38 | 39 | def rules(line, start): 40 | """ 处理特殊规则 """ 41 | if line[start] in numMath or line[start] in num_suffix: 42 | return proc_num_math(line, start) 43 | elif line[start] in numCn or line[start] in num_suffix: 44 | return proc_num_cn(line, start) 45 | else: 46 | return 1 47 | 48 | 49 | def genDict(path): 50 | """ 获取词典 """ 51 | f = codecs.open(path,'r','utf-8') 52 | contents = f.read() 53 | contents = contents.replace(u'\r', u'') 54 | contents = contents.replace(u'\n', u'') 55 | # 将内容逆置，以便进行逆向匹配 56 | contents = contents[::-1] 57 | # 将文件内容按空格分开 58 | mydict = contents.split(u' ') 59 | # 去除词典List中的重复 60 | newdict = list(set(mydict)) 61 | newdict.remove(u'') 62 | 63 | # 建立词典 64 | # key为词首字，value为以此字开始的词构成的List 65 | truedict = {} 66 | for item in newdict: 67 | if len(item)>0 and item[0] in truedict: 68 | value = truedict[item[0]] 69 | value.append(item) 70 | truedict[item[0]] = value 71 | else: 72 | truedict[item[0]] = [item] 73 | return truedict 74 | 75 | def print_unicode_list(uni_list): 76 | for item in uni_list: 77 | print item, 78 | 79 | def divideWords(mydict, sentence, maxlen): 80 | """ 81 | 根据词典对句子进行分词, 82 | 使用逆向匹配的算法，从右到左扫描，遇到最长的词， 83 | 就将它切下来，直到句子被分割完闭 84 | """ 85 | # 对句子逆置，以便用正向匹配算法进行实际的逆向处理 86 | sentence = sentence[::-1] 87 | ruleChar = [] 88 | ruleChar.extend(numCn) 89 | ruleChar.extend(numMath) 90 | result = [] 91 | start = 0 92 | senlen = len(sentence) 93 | while start < senlen: 94 | curword = sentence[start] 95 | wdlen = 1 96 | wdlen_rule = 1 97 | # 首先查看是否可以匹配特殊规则 98 | if curword in num_char or curword in num_suffix: 99 | wdlen_rule = rules(sentence, start) 100 | # 寻找以当前字开头的最长词 101 | if curword in mydict: 102 | wdlen = maxlen 103 | words = mydict[curword] 104 | while wdlen > 1: 105 | end = min(start+wdlen, senlen) 106 | if sentence[start:end] in words: 107 | break 108 | else: 109 | wdlen = wdlen - 1 110 | # 将新词使用[::-1]逆置，变为正常词序 111 | wdlen = max(wdlen_rule, wdlen) 112 | end = min(start+wdlen, senlen) 113 | result.append(sentence[start:end][::-1]) 114 | start = start + wdlen 115 | return result[::-1] 116 | 117 | def main(): 118 | args = sys.argv[1:] 119 | if len(args) < 3: 120 | print 'Usage: python ' + sys.argv[0] + ' dict_path test_path result_path' 121 | exit(-1) 122 | dict_path = args[0] 123 | test_path = args[1] 124 | result_path = args[2] 125 | 126 | dicts = genDict(dict_path) 127 | fr = codecs.open(test_path,'r','utf-8') 128 | test = fr.read() 129 | result = divideWords(dicts,test,5) 130 | fr.close() 131 | fw = codecs.open(result_path,'w','utf-8') 132 | for item in result: 133 | fw.write(item + ' ') 134 | fw.close() 135 | 136 | if __name__ == "__main__": 137 | main() 138 | -------------------------------------------------------------------------------- /src/cwsBMM_NRule.py: -------------------------------------------------------------------------------- 1 | #! /usr/bin/env python 2 | # -*- coding: utf-8 -*- 3 | 4 | # Author: 赵晓凯 5 | # Date: 2013-06-05 6 | # Email: minix007@foxmail.com 7 | 8 | # 使用BMM 进行中文分词 9 | 10 | import codecs 11 | import sys 12 | 13 | 14 | def genDict(path): 15 | """ 获取词典 """ 16 | f = codecs.open(path,'r','utf-8') 17 | contents = f.read() 18 | contents = contents.replace(u'\r', u'') 19 | contents = contents.replace(u'\n', u'') 20 | # 将内容逆置，以便进行逆向匹配 21 | contents = contents[::-1] 22 | # 将文件内容按空格分开 23 | mydict = contents.split(u' ') 24 | # 去除词典List中的重复 25 | newdict = list(set(mydict)) 26 | newdict.remove(u'') 27 | 28 | # 建立词典 29 | # key为词首字，value为以此字开始的词构成的List 30 | truedict = {} 31 | for item in newdict: 32 | if len(item)>0 and item[0] in truedict: 33 | value = truedict[item[0]] 34 | value.append(item) 35 | truedict[item[0]] = value 36 | else: 37 | truedict[item[0]] = [item] 38 | return truedict 39 | 40 | def print_unicode_list(uni_list): 41 | for item in uni_list: 42 | print item, 43 | 44 | def divideWords(mydict, sentence, maxlen): 45 | """ 46 | 根据词典对句子进行分词, 47 | 使用逆向匹配的算法，从右到左扫描，遇到最长的词， 48 | 就将它切下来，直到句子被分割完闭 49 | """ 50 | # 对句子逆置，以便用正向匹配算法进行实际的逆向处理 51 | sentence = sentence[::-1] 52 | result = [] 53 | start = 0 54 | senlen = len(sentence) 55 | while start < senlen: 56 | curword = sentence[start] 57 | wdlen = 1 58 | wdlen_rule = 1 59 | 60 | # 寻找以当前字开头的最长词 61 | if curword in mydict: 62 | wdlen = maxlen 63 | words = mydict[curword] 64 | while wdlen > 1: 65 | end = min(start+wdlen, senlen) 66 | if sentence[start:end] in words: 67 | break 68 | else: 69 | wdlen = wdlen - 1 70 | # 将新词使用[::-1]逆置，变为正常词序 71 | wdlen = max(wdlen_rule, wdlen) 72 | end = min(start+wdlen, senlen) 73 | result.append(sentence[start:end][::-1]) 74 | start = start + wdlen 75 | return result[::-1] 76 | 77 | def main(): 78 | args = sys.argv[1:] 79 | if len(args) < 3: 80 | print 'Usage: python' + sys.argv[0] + ' dict_path test_path result_path' 81 | exit(-1) 82 | dict_path = args[0] 83 | test_path = args[1] 84 | result_path = args[2] 85 | 86 | dicts = genDict(dict_path) 87 | fr = codecs.open(test_path,'r','utf-8') 88 | test = fr.read() 89 | result = divideWords(dicts,test,5) 90 | fr.close() 91 | fw = codecs.open(result_path,'w','utf-8') 92 | for item in result: 93 | fw.write(item + ' ') 94 | fw.close() 95 | 96 | if __name__ == "__main__": 97 | main() 98 | -------------------------------------------------------------------------------- /src/cwsFMM.py: -------------------------------------------------------------------------------- 1 | #! /usr/bin/env python 2 | # -*- coding: utf-8 -*- 3 | 4 | # Author: 赵晓凯 5 | # Date: 2013-06-05 6 | # Email: minix007@foxmail.com 7 | 8 | import codecs 9 | import sys 10 | 11 | # 由规则处理的一些特殊符号 12 | numMath = [u'0', u'1', u'2', u'3', u'4', u'5', u'6', u'7', u'8', u'9'] 13 | numMath_suffix = [u'.', u'%', u'亿', u'万', u'千', u'百', u'十', u'个'] 14 | numCn = [u'一', u'二', u'三', u'四', u'五', u'六', u'七', u'八', u'九', u'○', u'零'] 15 | numCn_suffix_date = [u'年', u'月', u'日'] 16 | numCn_suffix_unit = [u'亿', u'万', u'千', u'百', u'十', u'个'] 17 | special_char = [u'(', u')'] 18 | 19 | 20 | def proc_num_math(line, start): 21 | """ 处理句子中出现的数学符号 """ 22 | oldstart = start 23 | while line[start] in numMath or line[start] in numMath_suffix: 24 | start = start + 1 25 | if line[start] in numCn_suffix_date: 26 | start = start + 1 27 | return start - oldstart 28 | 29 | def proc_num_cn(line, start): 30 | """ 处理句子中出现的中文数字 """ 31 | oldstart = start 32 | while line[start] in numCn or line[start] in numCn_suffix_unit: 33 | start = start + 1 34 | if line[start] in numCn_suffix_date: 35 | start = start + 1 36 | return start - oldstart 37 | 38 | def rules(line, start): 39 | """ 处理特殊规则 """ 40 | if line[start] in numMath: 41 | return proc_num_math(line, start) 42 | elif line[start] in numCn: 43 | return proc_num_cn(line, start) 44 | 45 | def genDict(path): 46 | """ 获取词典 """ 47 | f = codecs.open(path,'r','utf-8') 48 | contents = f.read() 49 | contents = contents.replace(u'\r', u'') 50 | contents = contents.replace(u'\n', u'') 51 | # 将文件内容按空格分开 52 | mydict = contents.split(u' ') 53 | # 去除词典List中的重复 54 | newdict = list(set(mydict)) 55 | newdict.remove(u'') 56 | 57 | # 建立词典 58 | # key为词首字，value为以此字开始的词构成的List 59 | truedict = {} 60 | for item in newdict: 61 | if len(item)>0 and item[0] in truedict: 62 | value = truedict[item[0]] 63 | value.append(item) 64 | truedict[item[0]] = value 65 | else: 66 | truedict[item[0]] = [item] 67 | return truedict 68 | 69 | def print_unicode_list(uni_list): 70 | for item in uni_list: 71 | print item, 72 | 73 | def divideWords(mydict, sentence): 74 | """ 75 | 根据词典对句子进行分词, 76 | 使用正向匹配的算法，从左到右扫描，遇到最长的词， 77 | 就将它切下来，直到句子被分割完闭 78 | """ 79 | ruleChar = [] 80 | ruleChar.extend(numCn) 81 | ruleChar.extend(numMath) 82 | result = [] 83 | start = 0 84 | senlen = len(sentence) 85 | while start < senlen: 86 | curword = sentence[start] 87 | maxlen = 1 88 | # 首先查看是否可以匹配特殊规则 89 | if curword in numCn or curword in numMath: 90 | maxlen = rules(sentence, start) 91 | # 寻找以当前字开头的最长词 92 | if curword in mydict: 93 | words = mydict[curword] 94 | for item in words: 95 | itemlen = len(item) 96 | if sentence[start:start+itemlen] == item and itemlen > maxlen and itemlen <= 5: 97 | maxlen = itemlen 98 | 99 | result.append(sentence[start:start+maxlen]) 100 | start = start + maxlen 101 | return result 102 | 103 | def main(): 104 | args = sys.argv[1:] 105 | if len(args) < 3: 106 | print 'Usage: python ' + sys.argv[0] + ' dict_path test_path result_path' 107 | exit(-1) 108 | dict_path = args[0] 109 | test_path = args[1] 110 | result_path = args[2] 111 | 112 | dicts = genDict(dict_path) 113 | fr = codecs.open(test_path,'r','utf-8') 114 | test = fr.read() 115 | result = divideWords(dicts,test) 116 | fr.close() 117 | fw = codecs.open(result_path,'w','utf-8') 118 | for item in result: 119 | fw.write(item + ' ') 120 | fw.close() 121 | 122 | if __name__ == "__main__": 123 | main() 124 | -------------------------------------------------------------------------------- /src/cwsFMM_NRule.py: -------------------------------------------------------------------------------- 1 | #! /usr/bin/env python 2 | # -*- coding: utf-8 -*- 3 | 4 | # Author: 赵晓凯 5 | # Date: 2013-06-05 6 | # Email: minix007@foxmail.com 7 | 8 | import codecs 9 | import sys 10 | 11 | def genDict(path): 12 | """ 获取词典 """ 13 | f = codecs.open(path,'r','utf-8') 14 | contents = f.read() 15 | contents = contents.replace(u'\r', u'') 16 | contents = contents.replace(u'\n', u'') 17 | # 将文件内容按空格分开 18 | mydict = contents.split(u' ') 19 | # 去除词典List中的重复 20 | newdict = list(set(mydict)) 21 | newdict.remove(u'') 22 | 23 | # 建立词典 24 | # key为词首字，value为以此字开始的词构成的List 25 | truedict = {} 26 | for item in newdict: 27 | if len(item)>0 and item[0] in truedict: 28 | value = truedict[item[0]] 29 | value.append(item) 30 | truedict[item[0]] = value 31 | else: 32 | truedict[item[0]] = [item] 33 | return truedict 34 | 35 | def print_unicode_list(uni_list): 36 | for item in uni_list: 37 | print item, 38 | 39 | def divideWords(mydict, sentence): 40 | """ 41 | 根据词典对句子进行分词, 42 | 使用正向匹配的算法，从左到右扫描，遇到最长的词， 43 | 就将它切下来，直到句子被分割完闭 44 | """ 45 | result = [] 46 | start = 0 47 | senlen = len(sentence) 48 | while start < senlen: 49 | curword = sentence[start] 50 | maxlen = 1 51 | # 寻找以当前字开头的最长词 52 | if curword in mydict: 53 | words = mydict[curword] 54 | for item in words: 55 | itemlen = len(item) 56 | if sentence[start:start+itemlen] == item and itemlen > maxlen and itemlen <= 5: 57 | maxlen = itemlen 58 | result.append(sentence[start:start+maxlen]) 59 | start = start + maxlen 60 | return result 61 | 62 | def main(): 63 | args = sys.argv[1:] 64 | if len(args) < 3: 65 | print 'Usage: python ' + sys.argv[0] + ' dict_path test_path result_path' 66 | exit(-1) 67 | dict_path = args[0] 68 | test_path = args[1] 69 | result_path = args[2] 70 | 71 | dicts = genDict(dict_path) 72 | fr = codecs.open(test_path,'r','utf-8') 73 | test = fr.read() 74 | result = divideWords(dicts,test) 75 | fr.close() 76 | fw = codecs.open(result_path,'w','utf-8') 77 | for item in result: 78 | fw.write(item + ' ') 79 | fw.close() 80 | 81 | if __name__ == "__main__": 82 | main() 83 | -------------------------------------------------------------------------------- /src/cwsMaxEn-13f.py: -------------------------------------------------------------------------------- 1 | #! /usr/bin/env python 2 | # -*- coding: utf-8 -*- 3 | 4 | # Author: 赵晓凯 5 | # Date: 2013-06-05 6 | # Email: minix007@foxmail.com 7 | 8 | # 基于最大熵模型及字标注的分词工具，13特征版本 9 | 10 | import codecs 11 | import sys 12 | from maxent import MaxentModel 13 | 14 | def tag_training_set(training_file, tag_training_set_file): 15 | f = codecs.open(training_file,'r','utf-8') 16 | contents = f.read() 17 | contents = contents.replace(u'\r', u'') 18 | contents = contents.replace(u'\n', u'') 19 | # 将文件内容按空格分开 20 | words = contents.split(' ') 21 | print len(words) 22 | 23 | tag_words_list = [] 24 | i = 0 25 | for word in words: 26 | i += 1 27 | if (i % 100 == 0): tag_words_list.append(u'\r') 28 | if(len(word) == 0): 29 | continue 30 | if(len(word) == 1): 31 | tag_word = word + '/S' 32 | elif(len(word) == 2): 33 | tag_word = word[0] + '/B' + word[1] + '/E' 34 | elif(len(word) == 3): 35 | tag_word = word[0] + '/B' + word[1] + '/C' + word[2] + '/E' 36 | elif(len(word) == 4): 37 | tag_word = word[0] + '/B' + word[1] + '/C' + word[2] + '/D' + word[3] + '/E' 38 | else: 39 | tag_word = word[0] + '/B' + word[1] + '/C' + word[2] + '/D' 40 | mid_words = word[3:-1] 41 | for mid_word in mid_words: 42 | tag_word += (mid_word + '/M') 43 | tag_word += (word[-1] + '/E') 44 | 45 | tag_words_list.append(tag_word) 46 | 47 | tag_words = ''.join(tag_words_list) 48 | fw = codecs.open(tag_training_set_file, 'w', 'utf-8') 49 | fw.write(tag_words) 50 | fw.close() 51 | 52 | return (words,tag_words_list) 53 | 54 | def get_near_char(contents, i, times): 55 | words_len = len(contents) / times; 56 | if (i<0 or i >words_len-1): return '_' 57 | else: return contents[i*times] 58 | 59 | def get_near_tag(contents, i, times): 60 | words_len = len(contents) / times; 61 | if (i<0 or i >words_len-1): return '_' 62 | else: return contents[i*times+2] 63 | 64 | def isPu(char): 65 | punctuation = [u'，', u'。', u'？', u'！', u'；', u'－', u'、', u'—', u'（',u'）',u'《', u'》',u'：', 66 | u'“',u'”',u'’',u'‘'] 67 | if char in punctuation: 68 | return '1' 69 | else: 70 | return '0' 71 | 72 | def get_class(char): 73 | zh_num = [u'零',u'○',u'一', u'二',u'三',u'四',u'五',u'六',u'七',u'八',u'九',u'十',u'百',u'千',u'万'] 74 | ar_num = [u'0',u'1',u'2',u'3',u'4',u'5',u'6',u'7',u'8',u'9',u'.', 75 | u'０',u'１',u'２',u'３',u'４',u'５',u'６',u'７',u'８',u'９'] 76 | date = [u'日', u'年', u'月'] 77 | letter = ['a','b','c','d','e','f','g','h','i','j','k','l','m','n','o','p','q','r','s','t','u','v','w','x','y','z', 78 | 'A','B','C','D','E','F','G','H','I','J','K','L','M','N','O','P','Q','R','S','T','U','V','W','X','Y','Z'] 79 | if char in zh_num or char in ar_num: 80 | return '1' 81 | elif char in date: 82 | return '2' 83 | elif char in letter: 84 | return '3' 85 | else: 86 | return '4' 87 | 88 | 89 | def get_event(tag_file_path, event_file_path): 90 | f = codecs.open(tag_file_path,'r','utf-8') 91 | contents = f.read() 92 | contents = contents.replace(u'\r', u'') 93 | contents = contents.replace(u'\n', u'') 94 | words_len = len(contents)/3 95 | event_list = [] 96 | 97 | index = range(0,words_len) 98 | for i in index: 99 | pre_char = get_near_char(contents,i-1,3) 100 | pre_pre_char = get_near_char(contents,i-2,3) 101 | cur_char = get_near_char(contents,i,3) 102 | next_char = get_near_char(contents,i+1,3) 103 | next_next_char = get_near_char(contents,i+2,3) 104 | event_list.append( 105 | contents[i*3+2] + ' ' 106 | + 'C-2='+pre_pre_char + ' ' + 'C-1='+pre_char + ' ' 107 | + ' ' + 'C0='+cur_char + ' ' 108 | + 'C1='+next_char + ' ' + 'C2='+next_next_char + ' ' 109 | + 'C-2='+pre_pre_char+'C-1='+pre_char + ' ' 110 | + 'C-1='+pre_char+'C0='+cur_char + ' ' 111 | + 'C0='+cur_char+'C1='+next_char + ' ' 112 | + 'C1='+next_char+'C2='+next_next_char + ' ' 113 | + 'C-1='+pre_char+'C1='+next_char + ' ' 114 | + 'Pu='+isPu(cur_char) + ' ' 115 | + 'TC-2='+get_class(pre_pre_char)+'TC-1='+get_class(pre_char) 116 | + 'TC0='+get_class(cur_char)+'TC1='+get_class(next_char) 117 | + 'TC2='+get_class(next_next_char) + ' ' 118 | + 'T-1='+get_near_tag(contents,i-1,3) + ' ' 119 | + 'T-2='+get_near_tag(contents,i-2,3) 120 | + '\r') 121 | 122 | 123 | #events = ''.join(event_list) 124 | fw = codecs.open(event_file_path, 'w', 'utf-8') 125 | for event in event_list: 126 | fw.write(event) 127 | fw.close() 128 | 129 | return event_list 130 | 131 | def get_feature(test_file_path, feature_file_path): 132 | f = codecs.open(test_file_path,'r','utf-8') 133 | contents = f.read() 134 | contents_list = contents.split('\r\n') 135 | contents_list.remove('') 136 | contents_list.remove('') 137 | 138 | fw = codecs.open(feature_file_path, 'w', 'utf-8') 139 | for line in contents_list: 140 | words_len = len(line) 141 | feature_list = [] 142 | 143 | index = range(0,words_len) 144 | for i in index: 145 | pre_char = get_near_char(line,i-1,1) 146 | pre_pre_char = get_near_char(line,i-2,1) 147 | cur_char = get_near_char(line,i,1) 148 | next_char = get_near_char(line,i+1,1) 149 | next_next_char = get_near_char(line,i+2,1) 150 | feature_list.append( 151 | 'C-2='+pre_pre_char + ' ' + 'C-1='+pre_char + ' ' 152 | + 'C0='+cur_char + ' ' 153 | + 'C1='+next_char + ' ' + 'C2='+next_next_char + ' ' 154 | + 'C-2='+pre_pre_char+'C-1='+pre_char + ' ' 155 | + 'C-1='+pre_char+'C0='+cur_char + ' ' 156 | + 'C0='+cur_char+'C1='+next_char + ' ' 157 | + 'C1='+next_char+'C2='+next_next_char + ' ' 158 | + 'C-1='+pre_char+'C1='+next_char + ' ' 159 | + 'Pu='+isPu(cur_char) + ' ' 160 | + 'TC-2='+get_class(pre_pre_char)+'TC-1='+get_class(pre_char) 161 | + 'TC0='+get_class(cur_char)+'TC1='+get_class(next_char) 162 | + 'TC2='+get_class(next_next_char) + ' ' 163 | + '\r') 164 | 165 | for item in feature_list: 166 | fw.write(item) 167 | fw.write('split\r\n') 168 | 169 | fw.close() 170 | 171 | return feature_list 172 | 173 | def split_by_blank(line): 174 | line_list = [] 175 | line_len = len(line) 176 | i = 0 177 | while i < line_len: 178 | line_list.append(line[i]) 179 | i += 2 180 | 181 | return line_list 182 | 183 | def training(feature_file_path, trained_model_file, times): 184 | m = MaxentModel() 185 | fr = codecs.open(feature_file_path, 'r', 'utf-8') 186 | all_list = [] 187 | m.begin_add_event() 188 | for line in fr: 189 | line = line.rstrip() 190 | line_list = line.split(' ') 191 | str_list = [] 192 | for item in line_list: 193 | str_list.append(item.encode('utf-8')) 194 | all_list.append(str_list) 195 | m.add_event(str_list[1:], str_list[0], 1) 196 | m.end_add_event() 197 | print 'begin training' 198 | m.train(times, "lbfgs") 199 | print 'end training' 200 | m.save(trained_model_file) 201 | return all_list 202 | 203 | def max_prob(label_prob_list): 204 | max_prob = 0 205 | max_prob_label = '' 206 | for label_prob in label_prob_list: 207 | if label_prob[1] > max_prob: 208 | max_prob = label_prob[1] 209 | max_prob_label = label_prob[0] 210 | 211 | return max_prob_label 212 | 213 | def tag_test(test_feature_file, trained_model_file, tag_test_set_file): 214 | fr = codecs.open(test_feature_file, 'r', 'utf-8') 215 | fw = codecs.open(tag_test_set_file, 'w', 'utf-8') 216 | m = MaxentModel() 217 | m.load(trained_model_file) 218 | contents = fr.read() 219 | feature_list = contents.split('\r') 220 | feature_list.remove('\n') 221 | #return feature_list 222 | pre_tag = '_' 223 | pre_pre_tag = '_' 224 | for feature in feature_list: 225 | if (feature == 'split'): 226 | fw.write('\n\n\n') 227 | continue 228 | str_feature = [] 229 | u_feature = feature.split(' ') 230 | for item in u_feature: 231 | str_feature.append(item.encode('utf-8')) 232 | str_feature.append('T-1=' + pre_tag) 233 | str_feature.append('T-2=' + pre_pre_tag) 234 | label_prob_list = m.eval_all(str_feature) 235 | label = max_prob(label_prob_list) 236 | #print str_feature 237 | try: 238 | new_tag = str_feature[2].split('=')[1] + '/' + label 239 | except IndexError: 240 | print str_feature 241 | fw.write(new_tag.decode('utf-8')) 242 | pre_pre_tag = pre_tag 243 | pre_tag = label 244 | return feature_list 245 | 246 | def tag_to_words(tag_training_set_file, result_file): 247 | fr = codecs.open(tag_training_set_file, 'r', 'utf-8') 248 | fw = codecs.open(result_file, 'w', 'utf-8') 249 | 250 | contents = fr.read() 251 | words_len = len(contents)/3 252 | result = [] 253 | i = 0 254 | while (iwords_len-1): return '_' 57 | else: return contents[i*times] 58 | 59 | def get_near_tag(contents, i, times): 60 | words_len = len(contents) / times; 61 | if (i<0 or i >words_len-1): return '_' 62 | else: return contents[i*times+2] 63 | 64 | def isPu(char): 65 | punctuation = [u'，', u'。', u'？', u'！', u'；', u'－－', u'、', u'——', u'（',u'）',u'《', u'》',u'：', 66 | u'“',u'”',u'’',u'‘'] 67 | if char in punctuation: 68 | return '1' 69 | else: 70 | return '0' 71 | 72 | def get_class(char): 73 | zh_num = [u'零',u'○',u'一', u'二',u'三',u'四',u'五',u'六',u'七',u'八',u'九',u'十',u'百',u'千',u'万'] 74 | ar_num = [u'0',u'1',u'2',u'3',u'4',u'5',u'6',u'7',u'8',u'9',u'.', 75 | u'０',u'１',u'２',u'３',u'４',u'５',u'６',u'７',u'８',u'９'] 76 | date = [u'日', u'年', u'月'] 77 | letter = ['a','b','c','d','e','f','g','h','i','j','k','l','m','n','o','p','q','r','s','t','u','v','w','x','y','z', 78 | 'A','B','C','D','E','F','G','H','I','J','K','L','M','N','O','P','Q','R','S','T','U','V','W','X','Y','Z'] 79 | if char in zh_num or char in ar_num: 80 | return '1' 81 | elif char in date: 82 | return '2' 83 | elif char in letter: 84 | return '3' 85 | else: 86 | return '4' 87 | 88 | 89 | def get_event(tag_file_path, event_file_path): 90 | f = codecs.open(tag_file_path,'r','utf-8') 91 | contents = f.read() 92 | contents = contents.replace(u'\r', u'') 93 | contents = contents.replace(u'\n', u'') 94 | words_len = len(contents)/3 95 | event_list = [] 96 | 97 | index = range(0,words_len) 98 | for i in index: 99 | pre_char = get_near_char(contents,i-1,3) 100 | pre_pre_char = get_near_char(contents,i-2,3) 101 | cur_char = get_near_char(contents,i,3) 102 | next_char = get_near_char(contents,i+1,3) 103 | next_next_char = get_near_char(contents,i+2,3) 104 | event_list.append( 105 | contents[i*3+2] + ' ' 106 | + 'C-2='+pre_pre_char + ' ' + 'C-1='+pre_char + ' ' 107 | + ' ' + 'C0='+cur_char + ' ' 108 | + 'C1='+next_char + ' ' + 'C2='+next_next_char + ' ' 109 | + 'C-2='+pre_pre_char+'C-1='+pre_char + ' ' 110 | + 'C-1='+pre_char+'C0='+cur_char + ' ' 111 | + 'C0='+cur_char+'C1='+next_char + ' ' 112 | + 'C1='+next_char+'C2='+next_next_char + ' ' 113 | + 'C-1='+pre_char+'C1='+next_char + ' ' 114 | + 'C-2='+pre_pre_char+'C-1='+pre_char+'C0='+cur_char + ' ' 115 | + 'C-1='+pre_char+'C0='+cur_char + 'C1='+next_char + ' ' 116 | + 'C0='+cur_char + 'C1='+next_char + 'C2='+next_next_char + ' ' 117 | + 'Pu='+isPu(cur_char) + ' ' 118 | + 'TC-2='+get_class(pre_pre_char)+'TC-1='+get_class(pre_char) 119 | + 'TC0='+get_class(cur_char)+'TC1='+get_class(next_char) 120 | + 'TC2='+get_class(next_next_char) + ' ' 121 | + '\r') 122 | 123 | 124 | #events = ''.join(event_list) 125 | fw = codecs.open(event_file_path, 'w', 'utf-8') 126 | for event in event_list: 127 | fw.write(event) 128 | fw.close() 129 | 130 | return event_list 131 | 132 | def get_feature(test_file_path, feature_file_path): 133 | f = codecs.open(test_file_path,'r','utf-8') 134 | contents = f.read() 135 | contents_list = contents.split('\r\n') 136 | contents_list.remove('') 137 | contents_list.remove('') 138 | 139 | fw = codecs.open(feature_file_path, 'w', 'utf-8') 140 | for line in contents_list: 141 | words_len = len(line) 142 | feature_list = [] 143 | 144 | index = range(0,words_len) 145 | for i in index: 146 | pre_char = get_near_char(line,i-1,1) 147 | pre_pre_char = get_near_char(line,i-2,1) 148 | cur_char = get_near_char(line,i,1) 149 | next_char = get_near_char(line,i+1,1) 150 | next_next_char = get_near_char(line,i+2,1) 151 | feature_list.append( 152 | 'C-2='+pre_pre_char + ' ' + 'C-1='+pre_char + ' ' 153 | + 'C0='+cur_char + ' ' 154 | + 'C1='+next_char + ' ' + 'C2='+next_next_char + ' ' 155 | + 'C-2='+pre_pre_char+'C-1='+pre_char + ' ' 156 | + 'C-1='+pre_char+'C0='+cur_char + ' ' 157 | + 'C0='+cur_char+'C1='+next_char + ' ' 158 | + 'C1='+next_char+'C2='+next_next_char + ' ' 159 | + 'C-1='+pre_char+'C1='+next_char + ' ' 160 | + 'C-2='+pre_pre_char+'C-1='+pre_char+'C0='+cur_char + ' ' 161 | + 'C-1='+pre_char+'C0='+cur_char + 'C1='+next_char + ' ' 162 | + 'C0='+cur_char + 'C1='+next_char + 'C2='+next_next_char + ' ' 163 | + 'Pu='+isPu(cur_char) + ' ' 164 | + 'TC-2='+get_class(pre_pre_char)+'TC-1='+get_class(pre_char) 165 | + 'TC0='+get_class(cur_char)+'TC1='+get_class(next_char) 166 | + 'TC2='+get_class(next_next_char) + ' ' 167 | + '\r') 168 | 169 | for item in feature_list: 170 | fw.write(item) 171 | fw.write('split\r\n') 172 | 173 | fw.close() 174 | 175 | return feature_list 176 | 177 | def split_by_blank(line): 178 | line_list = [] 179 | line_len = len(line) 180 | i = 0 181 | while i < line_len: 182 | line_list.append(line[i]) 183 | i += 2 184 | 185 | return line_list 186 | 187 | def training(feature_file_path, trained_model_file, times): 188 | m = MaxentModel() 189 | fr = codecs.open(feature_file_path, 'r', 'utf-8') 190 | all_list = [] 191 | m.begin_add_event() 192 | for line in fr: 193 | line = line.rstrip() 194 | line_list = line.split(' ') 195 | str_list = [] 196 | for item in line_list: 197 | str_list.append(item.encode('utf-8')) 198 | all_list.append(str_list) 199 | m.add_event(str_list[1:], str_list[0], 1) 200 | m.end_add_event() 201 | print 'begin training' 202 | m.train(times, "lbfgs") 203 | print 'end training' 204 | m.save(trained_model_file) 205 | return all_list 206 | 207 | def max_prob(label_prob_list): 208 | max_prob = 0 209 | max_prob_label = '' 210 | for label_prob in label_prob_list: 211 | if label_prob[1] > max_prob: 212 | max_prob = label_prob[1] 213 | max_prob_label = label_prob[0] 214 | 215 | return max_prob_label 216 | 217 | def tag_test(test_feature_file, trained_model_file, tag_test_set_file): 218 | fr = codecs.open(test_feature_file, 'r', 'utf-8') 219 | fw = codecs.open(tag_test_set_file, 'w', 'utf-8') 220 | m = MaxentModel() 221 | m.load(trained_model_file) 222 | contents = fr.read() 223 | feature_list = contents.split('\r') 224 | feature_list.remove('\n') 225 | for feature in feature_list: 226 | if (feature == 'split'): 227 | fw.write('\n\n\n') 228 | continue 229 | str_feature = [] 230 | u_feature = feature.split(' ') 231 | for item in u_feature: 232 | str_feature.append(item.encode('utf-8')) 233 | label_prob_list = m.eval_all(str_feature) 234 | label = max_prob(label_prob_list) 235 | 236 | try: 237 | new_tag = str_feature[2].split('=')[1] + '/' + label 238 | except IndexError: 239 | print str_feature 240 | fw.write(new_tag.decode('utf-8')) 241 | pre_tag = label 242 | return feature_list 243 | 244 | def tag_to_words(tag_training_set_file, result_file): 245 | fr = codecs.open(tag_training_set_file, 'r', 'utf-8') 246 | fw = codecs.open(result_file, 'w', 'utf-8') 247 | 248 | contents = fr.read() 249 | words_len = len(contents)/3 250 | result = [] 251 | i = 0 252 | while (i) { 60 | chop; 61 | s/^\s*//; 62 | s/\s*$//; 63 | $dict{$_} = 1; 64 | } 65 | 66 | close(S); 67 | 68 | open (TRUTH, $ARGV[1]) or die "$ARGV[1]: $!\n"; 69 | open (TEST, $ARGV[2]) or die "$ARGV[2]: $!\n"; 70 | 71 | $Tot = $Del = $Ins = $Subst = $Truecount = $Testcount = 0; 72 | $RawRecall = $RawPrecision = 0; 73 | 74 | $linenum = 0; 75 | 76 | 77 | $IVMISSED = $OOVMISSED = $OOV = $IV = 0; 78 | 79 | $file1 = $ARGV[1]; 80 | $file2 = $ARGV[2]; 81 | $file1 =~ s=^/.*/==; 82 | $file2 =~ s=^/.*/==; 83 | 84 | while (defined($truth = ) && defined($test = )) { 85 | $truth =~ s/^\s*//; 86 | $test =~ s/^\s*//; 87 | $truth =~ s/\s*$//; 88 | $test =~ s/\s*$//; 89 | $truth =~ s/(\xe3\x80\x80)|(\xa1\x40)/ /g; 90 | $test =~ s/(\xe3\x80\x80)|(\xa1\x40)/ /g; 91 | $truth =~ s/ //g; 92 | $test =~ s/ //g; 93 | @truthwords = split /\s+/, $truth; 94 | @testwords = split /\s+/, $test; 95 | $truecount = scalar(@truthwords); 96 | $testcount = scalar(@testwords); 97 | ++$linenum; 98 | if ($truecount == 0) { 99 | if ($testcount > 0) { 100 | print STDERR "Warning: training is 0 but test is nonzero, possible misalignment at line $linenum.\n"; 101 | } 102 | next; 103 | } 104 | if ($testcount == 0) { 105 | print STDERR "Warning: No output in test data where there is in training data, line $linenum\n"; 106 | } 107 | open (T1, ">$tmp1") or die "Can't open $tmp1"; 108 | open (T2, ">$tmp2") or die "Can't open $tmp2"; 109 | foreach my $w (@truthwords) { print T1 "$w\n"; } 110 | foreach my $w (@testwords) {print T2 "$w\n";} 111 | close (T1); 112 | close (T2); 113 | open (P, "$diff -y $tmp1 $tmp2 |") 114 | or die "Can't open pipe.\n"; 115 | print "--$file1-------$file2----$linenum\n"; 116 | my $del = 0; 117 | my $ins = 0; 118 | my $subst = 0; 119 | my $rawrecall = 0; 120 | my $rawprecision = 0; 121 | while (

) { 122 | my $err = 0; 123 | if (/\s\|\s/) {$subst++ ; $err++; } 124 | elsif (/\s\>\s/) {$ins++ ; $err++; } 125 | elsif (/\s\<\s/) {$del++ ; $err++; } 126 | if (/^([^\s]+)\s/) { 127 | my $w = $1; 128 | if (!$dict{$w}) {++$OOV;} 129 | else {++$IV;} 130 | if (/^[^\s]+\s.*\s[\|\>\<]\s/) { 131 | if (!$dict{$w}) {++$OOVMISSED;} 132 | else {++$IVMISSED;} 133 | ++$rawrecall; 134 | } 135 | } 136 | if (/\s[\|\>\<]\s.*[^\s]$/) { ++$rawprecision; } 137 | print "$_"; 138 | } 139 | close (P); 140 | my $tot = $del + $ins + $subst; 141 | $Tot += $tot; 142 | $Del += $del; 143 | $Ins += $ins; 144 | $Subst += $subst; 145 | $Truecount += $truecount; 146 | $Testcount += $testcount; 147 | $rawrecall = $truecount - $rawrecall; 148 | $rawprecision = $testcount - $rawprecision; 149 | $RawRecall += $rawrecall; 150 | $RawPrecision += $rawprecision; 151 | $rawrecall = sprintf("%2.3f", $rawrecall/$truecount); 152 | $rawprecision = sprintf("%2.3f", $rawprecision/$testcount); 153 | print "INSERTIONS:\t$ins\n"; 154 | print "DELETIONS:\t$del\n"; 155 | print "SUBSTITUTIONS:\t$subst\n"; 156 | print "NCHANGE:\t$tot\n"; 157 | print "NTRUTH:\t$truecount\n"; 158 | print "NTEST:\t$testcount\n"; 159 | print "TRUE WORDS RECALL:\t$rawrecall\n"; 160 | print "TEST WORDS PRECISION:\t$rawprecision\n"; 161 | } 162 | 163 | close(TRUTH); 164 | close(TEST); 165 | unlink($tmp1); 166 | unlink($tmp2); 167 | 168 | print "=== SUMMARY:\n"; 169 | print "=== TOTAL INSERTIONS:\t$Ins\n"; 170 | print "=== TOTAL DELETIONS:\t$Del\n"; 171 | print "=== TOTAL SUBSTITUTIONS:\t$Subst\n"; 172 | print "=== TOTAL NCHANGE:\t$Tot\n"; 173 | print "=== TOTAL TRUE WORD COUNT:\t$Truecount\n"; 174 | print "=== TOTAL TEST WORD COUNT:\t$Testcount\n"; 175 | $RawRecall = $RawRecall/$Truecount; 176 | $RawPrecision = $RawPrecision/$Testcount; 177 | $beta = 1; 178 | $R = $RawRecall; 179 | $P = $RawPrecision; 180 | $F = (1 + $beta)*$P*$R/($beta * $P + $R); 181 | $F = sprintf("%2.3f", $F); 182 | $RawRecall = sprintf("%2.3f", $RawRecall); 183 | $RawPrecision = sprintf("%2.3f", $RawPrecision); 184 | print "=== TOTAL TRUE WORDS RECALL:\t$RawRecall\n"; 185 | print "=== TOTAL TEST WORDS PRECISION:\t$RawPrecision\n"; 186 | print "=== F MEASURE:\t$F\n"; 187 | if ($OOV > 0) { 188 | $OOVMISSED = sprintf("%2.3f", 1 - $OOVMISSED / $OOV); 189 | } 190 | else { 191 | $OOVMISSED = "--"; 192 | } 193 | $OOV = sprintf("%2.3f", $OOV / $Truecount); 194 | if ($IV > 0) { 195 | $IVMISSED = sprintf("%2.3f", 1 - $IVMISSED / $IV); 196 | } 197 | else { 198 | $IVMISSED = "--"; 199 | } 200 | print "=== OOV Rate:\t$OOV\n"; 201 | print "=== OOV Recall Rate:\t$OOVMISSED\n"; 202 | print "=== IV Recall Rate:\t$IVMISSED\n"; 203 | 204 | print "###\t$file2\t$Ins\t$Del\t$Subst\t$Tot\t$Truecount\t$Testcount\t$RawRecall\t$RawPrecision\t$F\t$OOV\t$OOVMISSED\t$IVMISSED\n"; 205 | exit(0); 206 | --------------------------------------------------------------------------------