Acquiring Predicate Paraphrases from News Tweets

Proceedings of the 6th Joint Conference on Lexical and Computational Semantics (*SEM 2017), pages 155–160, Vancouver, Canada, August 3-4, 2017. c⃝2017 Association for Computational Linguistics Acquiring Predicate Paraphrases from News Tweets Vered Shwartz Gabriel Stanovsky Computer Science Department, Bar-Ilan University, Ramat-Gan, Israel {vered1986, gabriel.satanovsky}@gmail.com, dagan@cs.biu.ac.il Ido Dagan Abstract We present a simple method for ever- growing extraction of predicate para- phrases from news headlines in Twitter. Analysis of the output of ten weeks of col- lection shows that the accuracy of para- phrases with different support levels is es- timated between 60-86%. We also demon- strate that our resource is to a large extent complementary to existing resources, pro- viding many novel paraphrases. Our re- source is publicly available, continuously expanding based on daily news. 1 Introduction Recognizing that various textual descriptions across multiple texts refer to the same event or action can benefit NLP applications such as rec- ognizing textual entailment (Dagan et al., 2013) and question answering. For example, to answer “when did the US Supreme Court approve same- sex marriage?” given the text “In June 2015, the Supreme Court ruled for same-sex marriage”, ap- prove and ruled for should be identified as describ- ing the same action. To that end, much effort has been devoted to identifying predicate paraphrases, some of which resulted in releasing resources of predicate entail- ment or paraphrases. Two main approaches were proposed for that matter; the first leverages the similarity in argument distribution across a large corpus between two predicates (e.g. [a]0 buy [a]1 / [a]0 acquire [a]1) (Lin and Pantel, 2001; Berant et al., 2010). The second approach exploits bilin- gual parallel corpora, extracting as paraphrases pairs of texts that were translated identically to for- eign languages (Ganitkevitch et al., 2013). While these methods have produced exhaustive resources which are broadly used by applications, [a]0 introduce [a]1 [a]0 welcome [a]1 [a]0 appoint [a]1 [a]0 to become [a]1 [a]0 die at [a]1 [a]0 pass away at [a]1 [a]0 hit [a]1 [a]0 sink to [a]1 [a]0 be investigate [a]1 [a]0 be probe [a]1 [a]0 eliminate [a]1 [a]0 slash [a]1 [a]0 announce [a]1 [a]0 unveil [a]1 [a]0 quit after [a]1 [a]0 resign after [a]1 [a]0 announce as [a]1 [a]0 to become [a]1 [a]0 threaten [a]1 [a]0 warn [a]1 [a]0 die at [a]1 [a]0 live until [a]1 [a]0 double down on [a]1 [a]0 stand by [a]1 [a]0 kill [a]1 [a]0 shoot [a]1 [a]0 approve [a]1 [a]0 pass [a]1 [a]0 would be cut under [a]1 [a]1 slash [a]0 seize [a]0 at [a]1 to grab [a]0 at [a]1 Table 1: A sample from the top-ranked predicate paraphrases. their accuracy is limited. Specifically, the first ap- proach may extract antonyms, that also have sim- ilar argument distribution (e.g. [a]0 raise to [a]1 / [a]0 fall to [a]1) while the second may conflate multiple senses of the foreign phrase. A third approach was proposed to harvest para- phrases from multiple mentions of the same event in news articles.1 This approach assumes that various redundant reports make different lexical choices …

Content truncated to 500 words