001/** 002 * Copyright (c) 2011, The University of Southampton and the individual contributors. 003 * All rights reserved. 004 * 005 * Redistribution and use in source and binary forms, with or without modification, 006 * are permitted provided that the following conditions are met: 007 * 008 * * Redistributions of source code must retain the above copyright notice, 009 * this list of conditions and the following disclaimer. 010 * 011 * * Redistributions in binary form must reproduce the above copyright notice, 012 * this list of conditions and the following disclaimer in the documentation 013 * and/or other materials provided with the distribution. 014 * 015 * * Neither the name of the University of Southampton nor the names of its 016 * contributors may be used to endorse or promote products derived from this 017 * software without specific prior written permission. 018 * 019 * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS" AND 020 * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED 021 * WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE 022 * DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR CONTRIBUTORS BE LIABLE FOR 023 * ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES 024 * (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; 025 * LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON 026 * ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT 027 * (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS 028 * SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. 029 */ 030package org.openimaj.text.nlp.sentiment.model.classifier; 031 032import java.util.ArrayList; 033import java.util.HashMap; 034import java.util.List; 035import java.util.Map.Entry; 036 037import org.openimaj.feature.DoubleFV; 038import org.openimaj.feature.FeatureExtractor; 039import org.openimaj.ml.annotation.AbstractAnnotator; 040import org.openimaj.ml.annotation.bayes.NaiveBayesAnnotator; 041 042/** 043 * {@link FeatureExtractor} that is suitable for {@link NaiveBayesAnnotator}. 044 * Should be initialized with training corpus of the machine learning 045 * {@link AbstractAnnotator} you are using. 046 * 047 * @author Laurence Willmore (lgw1e10@ecs.soton.ac.uk) 048 * 049 */ 050public class GeneralSentimentFeatureExtractor implements 051FeatureExtractor<DoubleFV, List<String>> 052{ 053 054 private ArrayList<String> vocabList; 055 private int wordOccuranceThresh = 50; 056 057 /** 058 * Construct with the training set. This is required to build a vocabulary. 059 * 060 * @param domainVocabularyCorpus 061 * list of tokenised corpus documents. 062 */ 063 public GeneralSentimentFeatureExtractor( 064 List<List<String>> domainVocabularyCorpus) 065 { 066 initialize(domainVocabularyCorpus); 067 } 068 069 /** 070 * Blank constructor. Will require initialize to be called at a later stage. 071 */ 072 public GeneralSentimentFeatureExtractor() { 073 074 } 075 076 /** 077 * Allows a new vocabulary to be constructed from a new corpus. 078 * 079 * @param domainVocabularyCorpus 080 * list of tokenised corpus documents. 081 */ 082 public void initialize(List<List<String>> domainVocabularyCorpus) { 083 final HashMap<String, Integer> vocab = new HashMap<String, Integer>(); 084 for (final List<String> doc : domainVocabularyCorpus) { 085 for (final String s : doc) { 086 Integer current = vocab.get(s); 087 if (current == null) 088 current = 0; 089 vocab.put(s, current + 1); 090 091 } 092 } 093 this.vocabList = new ArrayList<String>(); 094 for (final Entry<String, Integer> entry : vocab.entrySet()) { 095 if (entry.getValue() > wordOccuranceThresh) { 096 vocabList.add(entry.getKey()); 097 } 098 } 099 } 100 101 @Override 102 public DoubleFV extractFeature(List<String> tokens) { 103 final double[] vect = new double[vocabList.size()]; 104 for (int i = 0; i < vect.length; i++) { 105 vect[i] += 0.00001; 106 } 107 for (final String s : tokens) { 108 final int ind = vocabList.indexOf(s); 109 if (ind >= 0) 110 vect[ind] += 1; 111 } 112 final double[] vectNorm = new double[vocabList.size()]; 113 for (int i = 0; i < vect.length; i++) { 114 vectNorm[i] = vect[i] / tokens.size(); 115 } 116 return new DoubleFV(vect); 117 } 118 119}