diff options
Diffstat (limited to 'draklive/6/es/content/search/stemmers/fr_stemmer.js')
-rw-r--r-- | draklive/6/es/content/search/stemmers/fr_stemmer.js | 299 |
1 files changed, 299 insertions, 0 deletions
diff --git a/draklive/6/es/content/search/stemmers/fr_stemmer.js b/draklive/6/es/content/search/stemmers/fr_stemmer.js new file mode 100644 index 00000000..34f97431 --- /dev/null +++ b/draklive/6/es/content/search/stemmers/fr_stemmer.js @@ -0,0 +1,299 @@ +/* + * Author: Kasun Gajasinghe + * E-Mail: kasunbg AT gmail DOT com + * Date: 09.08.2010 + * + * usage: stemmer(word); + * ex: var stem = stemmer(foobar); + * Implementation of the stemming algorithm from http://snowball.tartarus.org/algorithms/french/stemmer.html + * + * LICENSE: + * + * Copyright (c) 2010, Kasun Gajasinghe. All rights reserved. + * + * Redistribution and use in source and binary forms, with or without modification, + * are permitted provided that the following conditions are met: + * + * 1. Redistributions of source code must retain the above copyright notice, + * this list of conditions and the following disclaimer. + * + * 2. Redistributions in binary form must reproduce the above copyright notice, + * this list of conditions and the following disclaimer in the documentation + * and/or other materials provided with the distribution. + * + * + * THIS SOFTWARE IS PROVIDED BY KASUN GAJASINGHE ''AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, + * INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A + * PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL KASUN GAJASINGHE BE LIABLE FOR ANY DIRECT, + * INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT + * LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR + * BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, + * STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE + * USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. + * + */ + +var stemmer = function(word){ +// Letters in French include the following accented forms, +// â à ç ë é ê è ï î ô û ù +// The following letters are vowels: +// a e i o u y â à ë é ê è ï î ô û ù + + word = word.toLowerCase(); + var oriWord = word; + word = word.replace(/qu/g, 'qU'); //have to perform first, as after the operation, capital U is not treated as a vowel + word = word.replace(/([aeiouyâàëéêèïîôûù])u([aeiouyâàëéêèïîôûù])/g, '$1U$2'); + word = word.replace(/([aeiouyâàëéêèïîôûù])i([aeiouyâàëéêèïîôûù])/g, '$1I$2'); + word = word.replace(/([aeiouyâàëéêèïîôûù])y/g, '$1Y'); + word = word.replace(/y([aeiouyâàëéêèïîôûù])/g, 'Y$1'); + + var rv=''; + var rvIndex = -1; + if(word.search(/^(par|col|tap)/) != -1 || word.search(/^[aeiouyâàëéêèïîôûù]{2}/) != -1){ + rv = word.substring(3); + rvIndex = 3; + } else { + rvIndex = word.substring(1).search(/[aeiouyâàëéêèïîôûù]/); + if(rvIndex != -1){ + rvIndex +=2; //+2 is to supplement the substring(1) used to find rvIndex + rv = word.substring(rvIndex); + } else { + rvIndex = word.length; + } + } + +// R1 is the region after the first non-vowel following a vowel, or the end of the word if there is no such non-vowel. +// R2 is the region after the first non-vowel following a vowel in R1, or the end of the word if there is no such non-vowel + var r1Index = word.search(/[aeiouyâàëéêèïîôûù][^aeiouyâàëéêèïîôûù]/); + var r1 = ''; + if (r1Index != -1) { + r1Index += 2; + r1 = word.substring(r1Index); + } else { + r1Index = word.length; + } + + var r2Index = -1; + var r2 = ''; + if (r1Index != -1) { + r2Index = r1.search(/[aeiouyâàëéêèïîôûù][^aeiouyâàëéêèïîôûù]/); + if (r2Index != -1) { + r2Index += 2; + r2 = r1.substring(r2Index); + r2Index += r1Index; + } else { + r2 = ''; + r2Index = word.length; + } + } + if (r1Index != -1 && r1Index < 3) { + r1Index = 3; + r1 = word.substring(r1Index); + } + + /* + Step 1: Standard suffix removal + */ + var a1Index = word.search(/(ance|iqUe|isme|able|iste|eux|ances|iqUes|ismes|ables|istes)$/); + var a2Index = word.search(/(atrice|ateur|ation|atrices|ateurs|ations)$/); + var a3Index = word.search(/(logie|logies)$/); + var a4Index = word.search(/(usion|ution|usions|utions)$/); + var a5Index = word.search(/(ence|ences)$/); + var a6Index = word.search(/(ement|ements)$/); + var a7Index = word.search(/(ité|ités)$/); + var a8Index = word.search(/(if|ive|ifs|ives)$/); + var a9Index = word.search(/(eaux)$/); + var a10Index = word.search(/(aux)$/); + var a11Index = word.search(/(euse|euses)$/); + var a12Index = word.search(/[^aeiouyâàëéêèïîôûù](issement|issements)$/); + var a13Index = word.search(/(amment)$/); + var a14Index = word.search(/(emment)$/); + var a15Index = word.search(/[aeiouyâàëéêèïîôûù](ment|ments)$/); + + if(a1Index != -1 && a1Index >= r2Index){ + word = word.substring(0,a1Index); + } else if(a2Index != -1 && a2Index >= r2Index){ + word = word.substring(0,a2Index); + var a2Index2 = word.search(/(ic)$/); + if(a2Index2 != -1 && a2Index2 >= r2Index){ + word = word.substring(0, a2Index2); //if preceded by ic, delete if in R2, + } else { //else replace by iqU + word = word.replace(/(ic)$/,'iqU'); + } + } else if(a3Index != -1 && a3Index >= r2Index){ + word = word.replace(/(logie|logies)$/,'log'); //replace with log if in R2 + } else if(a4Index != -1 && a4Index >= r2Index){ + word = word.replace(/(usion|ution|usions|utions)$/,'u'); //replace with u if in R2 + } else if(a5Index != -1 && a5Index >= r2Index){ + word = word.replace(/(ence|ences)$/,'ent'); //replace with ent if in R2 + } else if(a6Index != -1 && a6Index >= rvIndex){ + word = word.substring(0,a6Index); + if(word.search(/(iv)$/) >= r2Index){ + word = word.replace(/(iv)$/, ''); + if(word.search(/(at)$/) >= r2Index){ + word = word.replace(/(at)$/, ''); + } + } else if(word.search(/(eus)$/) != -1){ + var a6Index2 = word.search(/(eus)$/); + if(a6Index2 >=r2Index){ + word = word.substring(0, a6Index2); + } else if(a6Index2 >= r1Index){ + word = word.substring(0,a6Index2)+"eux"; + } + } else if(word.search(/(abl|iqU)$/) >= r2Index){ + word = word.replace(/(abl|iqU)$/,''); //if preceded by abl or iqU, delete if in R2, + } else if(word.search(/(ièr|Ièr)$/) >= rvIndex){ + word = word.replace(/(ièr|Ièr)$/,'i'); //if preceded by abl or iqU, delete if in R2, + } + } else if(a7Index != -1 && a7Index >= r2Index){ + word = word.substring(0,a7Index); //delete if in R2 + if(word.search(/(abil)$/) != -1){ //if preceded by abil, delete if in R2, else replace by abl, otherwise, + var a7Index2 = word.search(/(abil)$/); + if(a7Index2 >=r2Index){ + word = word.substring(0, a7Index2); + } else { + word = word.substring(0,a7Index2)+"abl"; + } + } else if(word.search(/(ic)$/) != -1){ + var a7Index3 = word.search(/(ic)$/); + if(a7Index3 != -1 && a7Index3 >= r2Index){ + word = word.substring(0, a7Index3); //if preceded by ic, delete if in R2, + } else { //else replace by iqU + word = word.replace(/(ic)$/,'iqU'); + } + } else if(word.search(/(iv)$/) != r2Index){ + word = word.replace(/(iv)$/,''); + } + } else if(a8Index != -1 && a8Index >= r2Index){ + word = word.substring(0,a8Index); + if(word.search(/(at)$/) >= r2Index){ + word = word.replace(/(at)$/, ''); + if(word.search(/(ic)$/) >= r2Index){ + word = word.replace(/(ic)$/, ''); + } else { word = word.replace(/(ic)$/, 'iqU'); } + } + } else if(a9Index != -1){ word = word.replace(/(eaux)/,'eau') + } else if(a10Index >= r1Index){ word = word.replace(/(aux)/,'al') + } else if(a11Index != -1 ){ + var a11Index2 = word.search(/(euse|euses)$/); + if(a11Index2 >=r2Index){ + word = word.substring(0, a11Index2); + } else if(a11Index2 >= r1Index){ + word = word.substring(0, a11Index2)+"eux"; + } + } else if(a12Index!=-1 && a12Index>=r1Index){ + word = word.substring(0,a12Index+1); //+1- amendment to non-vowel + } else if(a13Index!=-1 && a13Index>=rvIndex){ + word = word.replace(/(amment)$/,'ant'); + } else if(a14Index!=-1 && a14Index>=rvIndex){ + word = word.replace(/(emment)$/,'ent'); + } else if(a15Index!=-1 && a15Index>=rvIndex){ + word = word.substring(0,a15Index+1); + } + + /* Step 2a: Verb suffixes beginning i*/ + var wordStep1 = word; + var step2aDone = false; + if(oriWord == word.toLowerCase() || oriWord.search(/(amment|emment|ment|ments)$/) != -1){ + step2aDone = true; + var b1Regex = /([^aeiouyâàëéêèïîôûù])(îmes|ît|îtes|i|ie|ies|ir|ira|irai|iraIent|irais|irait|iras|irent|irez|iriez|irions|irons|iront|is|issaIent|issais|issait|issant|issante|issantes|issants|isse|issent|isses|issez|issiez|issions|issons|it)$/i; + if(word.search(b1Regex) >= rvIndex){ + word = word.replace(b1Regex,'$1'); + } + } + + /* Step 2b: Other verb suffixes*/ + if (step2aDone && wordStep1 == word) { + if (word.search(/(ions)$/) >= r2Index) { + word = word.replace(/(ions)$/, ''); + } else { + var b2Regex = /(é|ée|ées|és|èrent|er|era|erai|eraIent|erais|erait|eras|erez|eriez|erions|erons|eront|ez|iez)$/i; + if (word.search(b2Regex) >= rvIndex) { + word = word.replace(b2Regex, ''); + } else { + var b3Regex = /e(âmes|ât|âtes|a|ai|aIent|ais|ait|ant|ante|antes|ants|as|asse|assent|asses|assiez|assions)$/i; + if (word.search(b3Regex) >= rvIndex) { + word = word.replace(b3Regex, ''); + } else { + var b3Regex2 = /(âmes|ât|âtes|a|ai|aIent|ais|ait|ant|ante|antes|ants|as|asse|assent|asses|assiez|assions)$/i; + if (word.search(b3Regex2) >= rvIndex) { + word = word.replace(b3Regex2, ''); + } + } + } + } + } + + if(oriWord != word.toLowerCase()){ + /* Step 3 */ + var rep = ''; + if(word.search(/Y$/) != -1) { + word = word.replace(/Y$/, 'i'); + } else if(word.search(/ç$/) != -1){ + word = word.replace(/ç$/, 'c'); + } + } else { + /* Step 4 */ + //If the word ends s, not preceded by a, i, o, u, è or s, delete it. + if (word.search(/([^aiouès])s$/) >= rvIndex) { + word = word.replace(/([^aiouès])s$/, '$1'); + } + var e1Index = word.search(/ion$/); + if (e1Index >= r2Index && word.search(/[st]ion$/) >= rvIndex) { + word = word.substring(0, e1Index); + } else { + var e2Index = word.search(/(ier|ière|Ier|Ière)$/); + if (e2Index != -1 && e2Index >= rvIndex) { + word = word.substring(0, e2Index) + "i"; + } else { + if (word.search(/e$/) >= rvIndex) { + word = word.replace(/e$/, ''); //delete last e + } else if (word.search(/guë$/) >= rvIndex) { + word = word.replace(/guë$/, 'gu'); + } + } + } + } + + /* Step 5: Undouble */ + //word = word.replace(/(en|on|et|el|eil)(n|t|l)$/,'$1'); + word = word.replace(/(en|on)(n)$/,'$1'); + word = word.replace(/(ett)$/,'et'); + word = word.replace(/(el|eil)(l)$/,'$1'); + + /* Step 6: Un-accent */ + word = word.replace(/[éè]([^aeiouyâàëéêèïîôûù]+)$/,'e$1'); + word = word.toLowerCase(); + return word; +}; + +var eqOut = new Array(); +var noteqOut = new Array(); +var eqCount = 0; +/* +To test the stemming, create two arrays named "voc" and "COut" which are for vocabualary and the stemmed output. +Then add the vocabulary strings and output strings. This method will generate the stemmed output for "voc" and will +compare the output with COut. + (I used porter's voc and out files and did a regex to convert them to js objects. regex: /");\nvoc.push("/g . This + will add strings to voc array such that output would look like: voc.push("foobar"); ) drop me an email for any help. + */ +function testFr(){ + var start = new Date().getTime(); //execution time + eqCount = 0; + eqOut = new Array(); + noteqOut = new Array(); + for(var k=0;k<voc.length;k++){ + if(COut[k]==stemmer(voc[k])){ + eqCount++; + eqOut.push("v: "+voc[k]+" c: "+COut[k]); + } else { + noteqOut.push(voc[k]+", c: "+COut[k]+" s:"+stemmer(voc[k])); + } + } + var end = new Date().getTime(); //execution time + var time = end-start; + alert("equal count= "+eqCount+" out of "+voc.length+" words. time= "+time+" ms"); + //console.log("equal count= "+eqCount+" out of "+voc.length+" words. time= "+time+" ms"); +} + + |