Disambiguation of Verbal Shifters

Disambiguation of Verbal Shifters Michael Wiegand∗, Sylvette Loda∗, Josef Ruppenhofer† ∗Spoken Language Systems, Saarland University, †Institute for German Language ∗D-66123, Saarbr¨ucken, Germany, †D-68161, Mannheim, Germany michael.wiegand@lsv.uni-saarland.de, sylvette.loda@lsv.uni-saarland.de ruppenhofer@ids-mannheim.de Abstract Negation is an important contextual phenomenon that needs to be addressed in sentiment analysis. Next to common negation function words, such as not or none, there is also a considerably large class of negation content words, also referred to as shifters, such as the verbs diminish, reduce or reverse. However, many of these shifters are ambiguous. For instance, spoil as in spoil your chance reverses the polarity of the positive polar expression chance while in spoil your loved ones, no negation takes place. We present a supervised learning approach to disambiguating verbal shifters. Our approach takes into consideration various features, particularly generalization features. Keywords: sentiment analysis, negation modeling, word-sense disambiguation 1. Introduction Negation is one of the most central linguistic phenomena. Therefore, negation modeling is essential to various com- mon tasks in natural language processing, such as rela- tion extraction (Sanchez-Graillet and Poesio, 2007), recog- nition of textual entailment (Harabagiu et al., 2006) and particularly sentiment analysis (Wiegand et al., 2010). In the latter task, a negation typically reverses the polarity of polar expressions. For example, in (1), the negated positive polar expression pass conveys negative polarity. (1) Peter did [not [pass]+]− the exam. (2) Peter [failed to [pass]+]− the exam. So far, most research in negation modeling for sentiment analysis has focused on negation (function) words, such as the particle not or the adverbs no or never. However, among other word classes, particularly the content words, such as verbs, nouns and adjectives, there also exist words expressing negation. For example, in (2) the verb failed has a similar function as the negation word not in (1). These negation content words, which are also called shifters, are often excluded from discussion since there does not (yet) exist a commonly accepted resource with these expressions. Even though the frequency of a single negation function word is much higher than that of a shifter, the overall num- ber of shifters is significantly larger than those of negation function words. Schulder et al. (2017) identified 980 verbal shifters while the popular negation word lexicon proposed by Wilson et al. (2005) only includes 15 negation function words. Moreover, since content words are more ambiguous than function words, we also envisage shifters to be more ambiguous than negation function words. In this paper, we address the ambiguity of shifters. We se- lect a set of 20 ambiguous verbal shifters and try to disam- biguate them automatically. We focus on verbal shifters since it has been recently shown that a large amount of such verbs exist and they are important to polarity classi- fication (Schulder et al., 2017). Examples for ambiguous verbal shifters are clear, spoil, cloud or slump which in (3), (5), (7) and (9) convey negation while in (4), (6), (8) and (10) they do not. (3) The well-known actor was [cleared [of murder]−]+. (4) …

Content truncated to 500 words