Recognizing Complex Negation on Twitter

Recognizing Complex Negation on Twitter Junta Mizuno Canasai Kruengkrai Kiyonori Ohtake Chikara Hashimoto Kentaro Torisawa Julien Kloetzer National Institute of Information and Communications Technology Kyoto 619-0289, Japan {junta-m,canasai,kiyonori.ohtake,ch,torisawa,julien}@nict.go.jp Kentaro Inui Graduate School of Information Sceinces, Tohoku University Miyagi 980-8579, Japan inui@ecei.tohoku.ac.jp Abstract After the Great East Japan Earthquake in 2011, an abundance of false rumors were dis- seminated on Twitter that actually hindered rescue activities. This work presents a method for recognizing the negation of predicates on Twitter to find Japanese tweets that refute false rumors. We assume that the predicate “oc- cur” is negated in the sentence “The guy who tweeted that a nuclear explosion occurred has watched too many SF movies.” The challenge is in the treatment of such complex negation. We have to recognize a wide range of complex negation expressions such as “it is theoreti- cally impossible that...” and “The guy who... watched too many SF movies.” We tackle this problem using a combination of a supervised classifier and clusters of n-grams derived from large un-annotated corpora. The n-gram clus- ters give us a gain of about 22% in F-score for complex negations. 1 Introduction After the Great East Japan Earthquake in 2011, hun- dreds of false rumors were disseminated on Twitter. At the same time, many experts and other knowl- edgeable people posted tweets to refute such false rumors as “There is no truth to the rumor that a nu- clear explosion has occurred in Fukushima.” How- ever, since many people did not notice such refuta- tions, they retweeted the false rumors, inadvertently fueling the confusion and creating serious obstacles to rescue activities. This paper presents a method that recognizes the negation of predicates on Twitter to identify the tweets that refute false rumors1. Our proposed method uses a supervised learning method to judge whether a given predicate in a tweet is negated.. An important point here is that we have to deal with complex forms of negations to achieve our final goal; the detection of false rumors. Note that although our target data are Japanese tweets, we provide exam- ples in English for readability. S1 It is theoretically impossible that a nuclear ex- plosion occurred in Fukushima. S2 The guy who tweeted that a nuclear explosion occurred has watched too many SF movies. Both S1 and S2 refute that a nuclear explosion occurred. In other words, the predicate “occur” is negated, even though no negation words (e.g., “not”) are explicitly written in the sentences. Many sen- tences similar to the above examples were actually posted on Twitter to refute false rumors after the earthquake. In this paper, we categorize negation into two types: simple and complex. Given a predicate that is annotated as negation by a human annotator, its cat- egorization is done based on the following criteria. Simple negation If at least one of the words in the same phrase of the negated predicate ends with “ない nai, わけない wakenai, ぬ nu (all of which mean not)” we define this form as sim- …

Content truncated to 500 words