Recognizing stances in ideological on-line debates

From AcaWiki
Jump to: navigation, search

Citation: Swapna Somasundaran, Jaynce Wiebe Recognizing stances in ideological on-line debates. NAACL HLT 2010 workshop on Computational Approaches to Analysis and Generation of Emotion in Text, (RSS)
Internet Archive Scholar (search for fulltext): Recognizing stances in ideological on-line debates
Tagged: Computer Science (RSS) online argumentation (RSS), stance (RSS), NLP (RSS), natural language processing (RSS), argumentation mining (RSS)

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.

Selected References

Polarity

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, 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.


Retrieved from "https://acawiki.org/index.php?title=Recognizing_stances_in_ideological_on-line_debates&oldid=7305"