From Paraphrase Database to Compositional Paraphrase Model and Back

From Paraphrase Database to Compositional Paraphrase Model and Back John Wieting∗ Mohit Bansal† Kevin Gimpel† Karen Livescu† Dan Roth∗ ∗University of Illinois at Urbana-Champaign, Urbana, IL, 61801, USA {wieting2,danr}@illinois.edu †Toyota Technological Institute at Chicago, Chicago, IL, 60637, USA {mbansal,kgimpel,klivescu}@ttic.edu Abstract The Paraphrase Database (PPDB; Ganitke- vitch et al., 2013) is an extensive semantic re- source, consisting of a list of phrase pairs with (heuristic) confidence estimates. However, it is still unclear how it can best be used, due to the heuristic nature of the confidences and its necessarily incomplete coverage. We propose models to leverage the phrase pairs from the PPDB to build parametric paraphrase models that score paraphrase pairs more accurately than the PPDB’s internal scores while simul- taneously improving its coverage. They allow for learning phrase embeddings as well as im- proved word embeddings. Moreover, we in- troduce two new, manually annotated datasets to evaluate short-phrase paraphrasing mod- els. Using our paraphrase model trained using PPDB, we achieve state-of-the-art results on standard word and bigram similarity tasks and beat strong baselines on our new short phrase paraphrase tasks.1 1 Introduction Paraphrase detection2 is the task of analyzing two segments of text and determining if they have the same meaning despite differences in structure and wording. It is useful for a variety of NLP tasks like question answering (Rinaldi et al., 2003; Fader et al., 2013), semantic parsing (Berant and Liang, 2014), textual entailment (Bosma and Callison- Burch, 2007), and machine translation (Marton et al., 2009). 1We release our datasets, code, and trained models at http://web.engr.illinois.edu/˜wieting2/. 2See Androutsopoulos and Malakasiotis (2010) for a survey on approaches for detecting paraphrases. One component of many such systems is a para- phrase table containing pairs of text snippets, usu- ally automatically generated, that have the same meaning. The most recent work in this area is the Paraphrase Database (PPDB; Ganitkevitch et al., 2013), a collection of confidence-rated paraphrases created using the pivoting technique of Bannard and Callison-Burch (2005) over large parallel corpora. The PPDB is a massive resource, containing 220 million paraphrase pairs. It captures many short paraphrases that would be difficult to obtain us- ing any other resource. For example, the pair {we must do our utmost, we must make every effort} has little lexical overlap but is present in PPDB. The PPDB has recently been used for monolingual align- ment (Yao et al., 2013), for predicting sentence sim- ilarity (Bjerva et al., 2014), and to improve the cov- erage of FrameNet (Rastogi and Van Durme, 2014). Though already effective for multiple NLP tasks, we note some drawbacks of PPDB. The first is lack of coverage: to use the PPDB to compare two phrases, both must be in the database. The second is that PPDB is a nonparametric paraphrase model; the number of parameters (phrase pairs) grows with the size of the dataset used to build it. In practice, it can become unwieldy to work with as the size of the database increases. A third concern is that the confidence estimates …

Content truncated to 500 words