An Empirical Study on the Effect of Negation Words on Sentiment

Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics, pages 304–313, Baltimore, Maryland, USA, June 23-25 2014. c⃝2014 Association for Computational Linguistics An Empirical Study on the Effect of Negation Words on Sentiment Xiaodan Zhu, Hongyu Guo, Saif Mohammad and Svetlana Kiritchenko National Research Council Canada 1200 Montreal Road Ottawa, K1A 0R6, ON, Canada {Xiaodan.Zhu,Hongyu.Guo,Saif.Mohammad,Svetlana.Kiritchenko} @nrc-cnrc.gc.ca Abstract Negation words, such as no and not, play a fundamental role in modifying sentiment of textual expressions. We will refer to a negation word as the negator and the text span within the scope of the negator as the argument. Commonly used heuristics to estimate the sentiment of negated expres- sions rely simply on the sentiment of ar- gument (and not on the negator or the ar- gument itself). We use a sentiment tree- bank to show that these existing heuristics are poor estimators of sentiment. We then modify these heuristics to be dependent on the negators and show that this improves prediction. Next, we evaluate a recently proposed composition model (Socher et al., 2013) that relies on both the negator and the argument. This model learns the syntax and semantics of the negator’s ar- gument with a recursive neural network. We show that this approach performs bet- ter than those mentioned above. In ad- dition, we explicitly incorporate the prior sentiment of the argument and observe that this information can help reduce fitting er- rors. 1 Introduction Morante and Sporleder (2012) define negation to be “a grammatical category that allows the chang- ing of the truth value of a proposition”. Nega- tion is often expressed through the use of nega- tive signals or negators–words like isn’t and never, and it can significantly affect the sentiment of its scope. Understanding the impact of negation on sentiment is essential in automatic analysis of sentiment. The literature contains interesting re- search attempting to model and understand the behavior (reviewed in Section 2). For example, Figure 1: Effect of a list of common negators in modifying sentiment values in Stanford Senti- ment Treebank. The x-axis is s(⃗w), and y-axis is s(wn, ⃗w). Each dot in the figure corresponds to a text span being modified by (composed with) a negator in the treebank. The red diagonal line corresponds to the sentiment-reversing hypothesis that simply reverses the sign of sentiment values. a simple yet influential hypothesis posits that a negator reverses the sign of the sentiment value of the modified text (Polanyi and Zaenen, 2004; Kennedy and Inkpen, 2006). The shifting hypoth- esis (Taboada et al., 2011), however, assumes that negators change sentiment values by a constant amount. In this paper, we refer to a negation word as the negator (e.g., isn’t), a text span being mod- ified by and composed with a negator as the ar- gument (e.g., very good), and entire phrase (e.g., isn’t very good) as the negated phrase. The recently available Stanford Sentiment Tree- bank (Socher et al., 2013) renders manually anno- tated, real-valued sentiment scores for all phrases in parse …

Content truncated to 500 words