Recognizing stances in ideological on-line debates

{{Summary
 * title=Recognizing stances in ideological on-line debates
 * authors=Swapna Somasundaran, Jaynce Wiebe
 * tags=online argumentation, stance, NLP, natural language processing, argumentation mining
 * summary=Their research question is whether sentiment and expressions of opinion help classify ideological stances. Their approach is to create a supervised stance classifier.

The main novelty is the argumentation mining approach from social media. They use "arguing trigger expressions" (which are generally unigrams, bigrams, or trigrams) to detect arguments. They develop a lexicon with 3094 positive arguing expressions and 668 negative arguing trigger expressions, each associated with a probability.

First these expressions are detected in a sentence, then the associated probabilities are added (e.g. if multiple arguing expressions are found in a sentence). The remaining content words (nouns, verbs, adjectives, adverbs) are treated as the target, that we are arguing for or against.

Source topics analyzed
336 to 1186 posts are used for each of 6 topics.

Development

 * existence of God
 * healthcare

Experiments & analyses

 * gun rights
 * gay rights
 * abortion
 * creationism

Emphasis
Superlatives, imperatives, and sentiments are shown in the examples that show arguing.

Opinion Targets
The idea of opinion targets is key to this work and really interesting.

They manually annotate a corpus, creating "opinion-target pairs" which contain the opinion and what it is about. This paper uses "arguing" to mean linguistic subjectivity. They say: "in supporting their side, people not only express their sentiments, but they also argue about what is true... and about what should or should not be done...."

"Opinions by themselves may not be as informative as the combination of opinions and targets...It is by understanding what the opinion is about, that we can recognize the stance."

Experiments
Use SVM in Weka.

Controls

 * Distribution-based baseline (50% accurate)
 * Unigram (no explicit opinion info)
 * Useful in political domain per (Lin et al., 2006; Kim and Hovy, 2007)
 * "For example, a participant who chooses to speak about “child” and “life” in an abortion debate is more likely from an against-abortion side, while someone speaking about “woman”, “rape” and “choice” is more likely from a for-abortion stance."

Experiments using sentiment

 * Arguing System
 * Sentiment System
 * Arg+Sent System (using both arguing and sentiment features)

Arguing Features
The main arguing features are trigger expressions and modal verbs.

Trigger Expressions
They point to work by Wilson & Wiebe 2005 and Wilson 2007 on annotating the MPQA corpus version 2 for arguing subjectivity. Here are some sample trigger expressions from Table 2; more are listed in Table 3.

Positive arguing

 * actually
 * am convinced
 * bear witness
 * can only
 * has always

Negative arguing

 * certainly not
 * has never
 * not too
 * rather than
 * there is no

Sentiment-based features

 * Sentiment lexicon from Wilson et al. 2005
 * Positive
 * Negative
 * Neutral sentiment (e.g. “absolutely”, “amplify”, “believe”, and “think")

They treat the sentence polarity as unitary (either positive or negative but not both) and use the Vote and Flip algorithm (Choi and Cardie 2009).

Performance
Overall:
 * Sentiment
 * Unigram
 * Arg
 * Arg+Sent

However, in the Creationism domain, Unigram performs best.

Analysis

 * "There is an overlap between the content words used by Unigram, Arg+Sent and Arguing". Table 5 shows examples of these features for Unigram and Arg+Sent.

Polarity

 * Yejin Choi and Claire Cardie. 2009. Adapting a polarity lexicon using integer linear programming for domain-specific sentiment classification. In Proceedings of the 2009 Conference on Empirical Methods in Natural Language Processing, pages 590-598, Singapore, August. Association for Computational Linguistics.
 * Soo-Min Kim and Eduard Hovy. 2007. Crystal: Analyzing predictive opinions on the web. In Proceedings of the 2007 Joint Conference on Empirical Methods in Natural Language Processing and Computational Natural Language Learning (EMNLP-CoNLL), pages 1056-1064.

Stance
Wei-Hao Lin. 2006. Identifying perspectives at the document and sentence levels using statistical models. In Proceedings of the Human Language Technology Conference of the NAACL, Companion Volume: Doctoral Consortium, pages 227-230, New York City, USA, June. Association for Computational Linguistics.

Syntax
Stephan Greene and Philip Resnik. 2009. More than words: Syntactic packaging and implicit sentiment. In Proceedings of Human Language Technologies: The 2009 Annual Conference of the North American Chapter of the Association for Computational Linguistics, pages 503-511, Boulder, Colorado, June. Association for Computational Linguistics.

Related work on subjectivity

 * Swapna Somasundaran and Janyce Wiebe. 2009. Recognizing stances in online debates. In Proceedings of the Joint Conference of the 47th Annual Meeting of the ACL and the 4th International Joint Conference on Natural Language Processing of the AFNLP, pages 226-234, Suntec, Singapore, August. Association for Computational Linguistics.


 * Swapna Somasundaran, Josef Ruppenhofer, and Janyce Wiebe. 2007a. Detecting arguing and sentiment in meetings. In SIGdial Workshop on Discourse and Dialogue, Antwerp, Belgium, September.


 * Swapna Somasundaran, Theresa Wilson, Janyce Wiebe, and Veselin Stoyanov. 2007b. QA with attitude: Exploiting opinion type analysis for improving question answering in on-line discussions and the news. In International Conference on Weblogs and Social Media, Boulder, CO.


 * Swapna Somasundaran, Janyce Wiebe, and Josef Rup-penhofer. 2008. Discourse level opinion interpretation. In Proceedings of the 22nd International Conference on Computational Linguistics (Coling 2008), pages 801-808, Manchester, UK, August.


 * Theresa Wilson and Janyce Wiebe. 2005. Annotating attributions and private states. In Proceedings of ACL Workshop on Frontiers in Corpus Annotation II: Pie in the Sky.


 * Theresa Wilson, Janyce Wiebe, and Paul Hoffmann. 2005. Recognizing contextual polarity in phrase-level sentiment analysis. In hltemnlp2005, pages 347-354, Vancouver, Canada.

}}
 * Theresa Wilson. 2007. Fine-grained subjectivity and sentiment analysis: Recognizing the intensity, polarity, and attitudes of private states. Ph.D. thesis, Intelligent Systems Program, University of Pittsburgh.
 * journal=NAACL HLT 2010 workshop on Computational Approaches to Analysis and Generation of Emotion in Text,
 * subject=Computer Science