UNPKG

node-nlp

Version:

Library for NLU (Natural Language Understanding) done in Node.js

288 lines (238 loc) 10.5 kB
/* Copyright (c) 2018, Domingo Martín Mancera Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions: The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software. THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE. */ 'use strict'; var Stemmer = require('./stemmer-es'); // Inherit from the utility class in stemmer_es class PorterStemmer extends Stemmer { constructor() { super(); } isVowel(c) { var regex = /[aeiouáéíóú]/gi; return regex.test(c); } nextVowelPosition(word, start = 0) { var length = word.length; for(var position = start; position < length; position++){ if(this.isVowel(word[position])){ return position; } } return length; } nextConsonantPosition(word, start = 0) { var length = word.length; for(var position = start; position < length; position++){ if(!this.isVowel(word[position])){ return position; } } return length; } endsIn(word, suffix) { if(word.length < suffix.length){ return false; } return (word.slice(-suffix.length) === suffix); } endsInArr(word, suffixes) { var matches = []; for(var i in suffixes) { if(this.endsIn(word, suffixes[i])){ matches.push(suffixes[i]); } } var longest = matches.sort(function (a, b) { return b.length - a.length; })[0]; if(longest){ return longest } else { return ''; } } removeAccent(word) { var accentedVowels = ['á', 'é', 'í', 'ó', 'ú']; var vowels = ['a', 'e', 'i', 'o', 'u']; for(var i in accentedVowels){ word = word.replace(accentedVowels[i], vowels[i]); } return word; } stem(word) { var length = word.length; word.toLowerCase(); if(length < 2){ return this.removeAccent(word); } var r1, r2, rv; r1 = length; r2 = length; rv = length; // R1 is the region after the first non-vowel following a vowel, or is the null region // at the end of the word if there is no such non-vowel. for(var i = 0; i < (length - 1) && r1 == length; i++){ if(this.isVowel(word[i]) && !this.isVowel(word[i + 1])){ r1 = i + 2; } } // R2 is the region after the first non-vowel following a vowel in R1, // or is the null region at the end of the word if there is no such non-vowel. for(var i = r1; i < (length - 1) && r2 == length; i++){ if(this.isVowel(word[i]) && !this.isVowel(word[i + 1])){ r2 = i + 2; } } if(length > 3){ if(!this.isVowel(word[1])){ rv = this.nextVowelPosition(word, 2) + 1; } else if(this.isVowel(word[0]) && this.isVowel(word[1])){ rv = this.nextConsonantPosition(word, 2) + 1; } else { rv = 3; } } var r1Text = word.slice(r1); var r2Text = word.slice(r2); var rvText = word.slice(rv); var originalWord = word; // Step 0: Attached pronoun var pronounSuffix = ['me', 'se', 'sela', 'selo', 'selas', 'selos', 'la', 'le', 'lo', 'las', 'les', 'los', 'nos']; var pronounSuffixPre1 = ['iéndo', 'ándo', 'ár', 'ér', 'ír']; var pronounSuffixPre2 = ['iendo', 'ando', 'ar', 'er', 'ir']; var suffix = this.endsInArr(word, pronounSuffix); if(suffix != ''){ var preSuffix = this.endsInArr(rvText.slice(0, -suffix.length), pronounSuffixPre1); if(preSuffix != ''){ word = this.removeAccent(word.slice(0, -suffix.length)); } else { preSuffix = this.endsInArr(rvText.slice(0, -suffix.length), pronounSuffixPre2); if(preSuffix != '' || (this.endsIn(word.slice(0, -suffix.length), 'uyendo'))){ word = word.slice(0, -suffix.length); } } } if(word != originalWord){ r1Text = word.slice(r1); r2Text = word.slice(r2); rvText = word.slice(rv); } var wordAfter0 = word; if(( suf = this.endsInArr(r2Text, ['anza', 'anzas', 'ico', 'ica', 'icos', 'icas', 'ismo', 'ismos', 'able', 'ables', 'ible', 'ibles', 'ista', 'istas', 'oso', 'osa', 'osos', 'osas', 'amiento', 'amientos', 'imiento', 'imientos'])) != '') { word = word.slice(0, -suf.length); } else if((suf = this.endsInArr(r2Text, ['icadora', 'icador', 'icación', 'icadoras', 'icadores', 'icaciones', 'icante', 'icantes', 'icancia', 'icancias', 'adora', 'ador', 'ación', 'adoras', 'adores', 'aciones', 'ante', 'antes', 'ancia', 'ancias'])) != '') { word = word.slice(0, -suf.length); } else if((suf = this.endsInArr(r2Text, ['logía', 'logías'])) != ''){ word = word.slice(0, -suf.length) + 'log'; } else if((suf = this.endsInArr(r2Text, ['ución', 'uciones'])) != ''){ word = word.slice(0, -suf.length) + 'u'; } else if((suf = this.endsInArr(r2Text, ['encia', 'encias'])) != ''){ word = word.slice(0, -suf.length) + 'ente'; } else if((suf = this.endsInArr(r2Text, ['ativamente', 'ivamente', 'osamente', 'icamente', 'adamente'])) != ''){ word = word.slice(0, -suf.length); } else if((suf = this.endsInArr(r1Text, ['amente'])) != ''){ word = word.slice(0, -suf.length); } else if((suf = this.endsInArr(r2Text, ['antemente', 'ablemente', 'iblemente', 'mente'])) != ''){ word = word.slice(0, -suf.length); } else if((suf = this.endsInArr(r2Text, ['abilidad', 'abilidades', 'icidad', 'icidades', 'ividad', 'ividades', 'idad', 'idades'])) != ''){ word = word.slice(0, -suf.length); } else if((suf = this.endsInArr(r2Text, ['ativa', 'ativo', 'ativas', 'ativos', 'iva', 'ivo', 'ivas', 'ivos'])) != ''){ word = word.slice(0, -suf.length); } if(word != wordAfter0){ r1Text = word.slice(r1); r2Text = word.slice(r2); rvText = word.slice(rv); } var wordAfter1 = word; if(wordAfter0 === wordAfter1){ // Do step 2a if no ending was removed by step 1. var suf = this.endsInArr(rvText, ['ya', 'ye', 'yan', 'yen', 'yeron', 'yendo', 'yo', 'yó', 'yas', 'yes', 'yais', 'yamos']); if(suf != '' && (word.slice(-suf.length - 1, -suf.length) == 'u')){ word = word.slice(0, -suf.length); } if(word != wordAfter1){ r1Text = word.slice(r1); r2Text = word.slice(r2); rvText = word.slice(rv); } var wordAfter2a = word; // Do Step 2b if step 2a was done, but failed to remove a suffix. if (wordAfter2a == wordAfter1) { if((suf = this.endsInArr(rvText, ['arían', 'arías', 'arán', 'arás', 'aríais', 'aría', 'aréis', 'aríamos', 'aremos', 'ará', 'aré', 'erían', 'erías', 'erán', 'erás', 'eríais', 'ería', 'eréis', 'eríamos', 'eremos', 'erá', 'eré', 'irían', 'irías', 'irán', 'irás', 'iríais', 'iría', 'iréis', 'iríamos', 'iremos', 'irá', 'iré', 'aba', 'ada', 'ida', 'ía', 'ara', 'iera', 'ad', 'ed', 'id', 'ase', 'iese', 'aste', 'iste', 'an', 'aban', 'ían', 'aran', 'ieran', 'asen', 'iesen', 'aron', 'ieron', 'ado', 'ido', 'ando', 'iendo', 'ió', 'ar', 'er', 'ir', 'as', 'abas', 'adas', 'idas', 'ías', 'aras', 'ieras', 'ases', 'ieses', 'ís', 'áis', 'abais', 'íais', 'arais', 'ierais', ' aseis', 'ieseis', 'asteis', 'isteis', 'ados', 'idos', 'amos', 'ábamos', 'íamos', 'imos', 'áramos', 'iéramos', 'iésemos', 'ásemos'])) != '') { word = word.slice(0, -suf.length); }else if((suf = this.endsInArr(rvText, ['en', 'es', 'éis', 'emos'])) != '') { word = word.slice(0, -suf.length); if(this.endsIn(word, 'gu')){ word = word.slice(0, -1); } } } } r1Text = word.slice(r1); r2Text = word.slice(r2); rvText = word.slice(rv); if ((suf = this.endsInArr(rvText, ['os', 'a', 'o', 'á', 'í', 'ó'])) != '') { word = word.slice(0, -suf.length); } else if ((this.endsInArr(rvText , ['e','é'])) != '') { word = word.slice(0, -1); rvText = word.slice(rv); if (this.endsIn(rvText, 'u') && this.endsIn(word, 'gu')) { word = word.slice(0, -1); } } return this.removeAccent(word); } } module.exports = new PorterStemmer();