Multilingual Affect Polarity and Valence Prediction in Metaphor-Rich Texts

Proceedings of the 51st Annual Meeting of the Association for Computational Linguistics, pages 682–691, Sofia, Bulgaria, August 4-9 2013. c⃝2013 Association for Computational Linguistics Multilingual Affect Polarity and Valence Prediction in Metaphor-Rich Texts Zornitsa Kozareva USC Information Sciences Institute 4676 Admiralty Way Marina del Rey, CA 90292-6695 kozareva@isi.edu Abstract Metaphor is an important way of convey- ing the affect of people, hence understand- ing how people use metaphors to convey affect is important for the communication between individuals and increases cohe- sion if the perceived affect of the con- crete example is the same for the two in- dividuals. Therefore, building computa- tional models that can automatically iden- tify the affect in metaphor-rich texts like “The team captain is a rock.”, “Time is money.”, “My lawyer is a shark.” is an important challenging problem, which has been of great interest to the research com- munity. To solve this task, we have collected and manually annotated the affect of metaphor-rich texts for four languages. We present novel algorithms that integrate triggers for cognitive, affective, perceptual and social processes with stylistic and lex- ical information. By running evaluations on datasets in English, Spanish, Russian and Farsi, we show that the developed af- fect polarity and valence prediction tech- nology of metaphor-rich texts is portable and works equally well for different lan- guages. 1 Introduction Metaphor is a figure of speech in which a word or phrase that ordinarily designates one thing is used to designate another, thus making an implicit comparison (Lakoff and Johnson, 1980; Martin, 1988; Wilks, 2007). For instance, in “My lawyer is a shark” the speaker may want to communicate that his/her lawyer is strong and aggressive, and that he will attack in court and persist until the goals are achieved. By using the metaphor, the speaker ac- tually conveys positive affect because having an aggressive lawyer is good if one is being sued. There has been a substantial body of work on metaphor identification and interpretation (Wilks, 2007; Shutova et al., 2010). However, in this paper we focus on an equally interesting, chal- lenging and important problem, which concerns the automatic identification of affect carried by metaphors. Building such computational mod- els is important to understand how people use metaphors to convey affect and how affect is ex- pressed using metaphors. The existence of such models can be also used to improve the communi- cation between individuals and to make sure that the speakers perceived the affect of the concrete metaphor example in the same way. The questions we address in this paper are: “How can we build computational models that can identify the polarity and valence associated with metaphor-rich texts?” and “Is it possible to build such automatic models for multiple languages?”. Our main contributions are: • We have developed multilingual metaphor- rich datasets in English, Spanish, Russian and Farsi that contain annotations of the Positive and Negative polarity and the valence (from −3 to +3 scale) corresponding to the inten- sity of the affect conveyed in …

Content truncated to 500 words