04 10 2014 Hyun Geun Soo Bo Pang and Lillian Lee 2004 ACL04 Outline Introduction Method Evaluation Framework Experimental Results Conclusions Intro Sentiment analysis Identify the view point underlying a text span ID: 414433
Download Presentation The PPT/PDF document "A Sentimental Education: Sentiment Analy..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
A Sentimental Education: Sentiment Analysis Using Subjectivity Summarization Based on Minimum Cuts
04 10, 2014Hyun Geun Soo
Bo Pang and Lillian Lee (2004) ACL-04Slide2
OutlineIntroductionMethod
Evaluation FrameworkExperimental ResultsConclusionsSlide3
IntroSentiment analysisIdentify the view point underlying a text spanSentiment polarity
E.g. classifying a movie review “thumbs up” “thumbs down”In this paper,Novel maching learning methodMinimum cuts in graphsSlide4
IntroPrevious Document polarity classification focused on selecting indicative lexical feature(e.g. good), classifying the number of such features
In this paper,1) label the sentences in the document as either subjective or objective and discarding latter2) apply a standard machine learning classifier to the resulting extractPrevent, irrelevant or potentially misleading textE.g. “The protagonist tries to protect her good name”Summary of the sentiment-oriented content of the documentSlide5
OutlineIntroductionMethod
Evaluation FrameworkExperimental ResultsConclusionsSlide6
ArchitectureSVM( Support vector machines )… – default polarity classifiersRemoving objective sentence (e.g. plot summaries) – subjectivity detectorSlide7
Context and Subjectivity DetectionStandard classification algorithm apply on each sentence in isolationNaïve Bayes or SVM classifiers label each test item in isolation
to specify that two particular sentences should ideally receive the same subjectivity label but not state which label this should beModeling proximity relationshipsShare the same subjectivity status, other things being equalOur method, minimum cutsConcerned with physical proximity between the items to be classifiedSlide8
Cut-based classificationSlide9
Cut-based classificationMinimum-cut practical advantagesModel item specific and pair-wise information independentlyCan use maximum-flow algorithms with polynomial asymptotic running times
Other graph-partitioning problems are NP-complete Slide10
OutlineIntroductionMethod
Evaluation FrameworkExperimental ResultsConclusionsSlide11
Evaluation FrameworkClassifying movie reviews as either positive or negativeProviding polarity information about reviews is a useful serviceMovie reviews are apparently harder to classify than reviews of other product
The correct label can be extracted automatically from rating informationPolarity dataset1000 positive and 1000 negative reviewsDefault polarity classifiers – SVMs, NBSubjectivity dataset 5000 movie review snippets and 5000 sentences from plot summaries Subjectivity detectorsBasic sentence level subjectivity detectorCut based subjectivity detectorSlide12
Evaluation FrameworkSubjectivity detectorsSource s , sink t = class of subjective and objectiveInd
(s) = (denote Naïve Bayes’ estimate of the probility that sentence s is subjective). Slide13
OutlineIntroductionMethod
Evaluation FrameworkExperimental ResultsConclusionsSlide14
Experimental resultsTen fold cross validationSubjectivity extraction produces effective summaries of document sentiment
Basic subjectivity extractionNaïve Bayes and SVMsIncorporating context informationNaïve Bayes + min-cut and SVMs + min-cutSlide15
Basic subjectivity extractionNaïve Bayes and SVMs can be trained on our subjectivity dataset
Naïve Bayes subjectivity detector + Naïve Bayes polarity classifier82% -> 86% improve than no extractionN most subjective sentencesLast N sentencesFirst N sentencesLeast subjective N sentencesSlide16
Experimental resultsSlide17
Experimental resultsSlide18
OutlineIntroductionMethod
Evaluation FrameworkExperimental ResultsConclusionsSlide19
ConclusionShowing that subjectivity detection can compress reviews into much shorter extracts that still retain polarity information at a level comparable to that of the full review
For NB classifier, Extraction is not only shorter but also cleaner representationsUtilizing contextual information via this framework can lead to statistically significant improvement in polarity classification accuracy