+/-EffectWordNet: Sense-level Lexicon Acquisition for Opinion Inference

Proceedings of the 2014 Conference on Empirical Methods in Natural Language Processing (EMNLP), pages 1181–1191, October 25-29, 2014, Doha, Qatar. c⃝2014 Association for Computational Linguistics +/-EffectWordNet: Sense-level Lexicon Acquisition for Opinion Inference Yoonjung Choi and Janyce Wiebe Department of Computer Science University of Pittsburgh yjchoi, wiebe@cs.pitt.edu Abstract Recently, work in NLP was initiated on a type of opinion inference that arises when opinions are expressed toward events which have positive or negative effects on entities (+/-effect events). This paper addresses methods for creating a lexicon of such events, to support such work on opinion inference. Due to significant sense ambiguity, our goal is to develop a sense-level rather than word-level lexicon. To maximize the effectiveness of different types of information, we combine a graph-based method using WordNet1 relations and a standard classifier using gloss information. A hybrid between the two gives the best results. Further, we provide evidence that the model is an effective way to guide manual annotation to find +/-effect senses that are not in the seed set. 1 Introduction Opinion mining (or sentiment analysis) identifies positive or negative opinions in many kinds of texts such as reviews, blogs, and news articles. It has been exploited in many application areas such as review mining, election analysis, and infor- mation extraction. While most previous research focusses on explicit opinion expressions, recent work addresses a type of opinion inference that arises when opinions are expressed toward events which have positive or negative effects on enti- ties (Deng et al., 2013; Deng and Wiebe, 2014). We call such events +/-effect events.2 Deng and Wiebe (2014) show how sentiments toward one 1WordNet 3.0, http://wordnet.princeton.edu/ 2While the term goodFor/badFor is used in previous pa- pers (Deng et al., 2013; Deng and Wiebe, 2014; Deng et al., 2014), we have since decided that +/-effect is a better term. entity may be propagated to other entities via opinion inference rules. They give the following example: (1) The bill would curb skyrocketing health care costs. The writer expresses an explicit negative senti- ment (by skyrocketing) toward the object (health care costs). The event, curb, has a negative effect on costs, since they are reduced. We can reason that the writer is positive toward the event because it has a negative effect on costs, toward which the writer is negative. From there, we can reason that the writer is positive toward the bill, since it is the agent of the positive event. Deng and Wiebe (2014) show that such inferences may be exploited to significantly improve explicit sentiment analy- sis systems. However, to achieve its results, the system de- veloped by Deng and Wiebe (2014) requires that all instances of +/-effect events in the corpus be manually provided as input. For the system to be fully automatic, it needs to be able to recog- nize +/-effect events automatically. This paper addresses methods for creating lexicons of such events, to support such work on opinion inference. We have discovered that there is significant sense ambiguity, meaning that words …

Content truncated to 500 words