POLY: Mining Relational Paraphrases from Multilingual Sentences

Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing, pages 2183–2192, Austin, Texas, November 1-5, 2016. c⃝2016 Association for Computational Linguistics POLY: Mining Relational Paraphrases from Multilingual Sentences Adam Grycner Max-Planck Institute for Informatics Saarland Informatics Campus Building E1.4, 66123 Saarbr¨ucken, Germany agrycner@mpi-inf.mpg.de Gerhard Weikum Max-Planck Institute for Informatics Saarland Informatics Campus Building E1.4, 66123 Saarbr¨ucken, Germany weikum@mpi-inf.mpg.de Abstract Language resources that systematically orga- nize paraphrases for binary relations are of great value for various NLP tasks and have re- cently been advanced in projects like PATTY, WiseNet and DEFIE. This paper presents a new method for building such a resource and the resource itself, called POLY. Starting with a very large collection of multilingual sentences parsed into triples of phrases, our method clusters relational phrases using prob- abilistic measures. We judiciously leverage fine-grained semantic typing of relational ar- guments for identifying synonymous phrases. The evaluation of POLY shows significant im- provements in precision and recall over the prior works on PATTY and DEFIE. An ex- trinsic use case demonstrates the benefits of POLY for question answering. 1 Introduction Motivation. Information extraction from text typi- cally yields relational triples: a binary relation along with its two arguments. Often the relation is ex- pressed by a verb phrase, and the two arguments are named entities. We refer to the surface form of the relation in a triple as a relational phrase. Reposito- ries of relational phrases are an asset for a variety of tasks, including information extraction, textual en- tailment, and question answering. This paper presents a new method for systemat- ically organizing a large set of such phrases. We aim to construct equivalence classes of synonymous phrases, analogously to how WordNet organizes unary predicates as noun-centric synsets (aka. se- mantic types). For example, the following relational phrases should be in the same equivalence class: sings in, is vocalist in, voice in denoting a relation between a musician and a song. State of the Art and its Limitations. Starting with the seminal work on DIRT (Lin and Pantel, 2001), there have been various attempts on build- ing comprehensive resources for relational phrases. Recent works include PATTY (Nakashole et al., 2012), WiseNet (Moro and Navigli, 2012) and DE- FIE (Bovi et al., 2015). Out of these DEFIE is the cleanest resource. However, the equivalence classes tend to be small, prioritizing precision over recall. On the other hand, PPDB (Ganitkevitch et al., 2013) offers the largest repository of paraphrases. How- ever, the paraphrases are not relation-centric and they are not semantically typed. So it misses out on the opportunity of using types to distinguish iden- tical phrases with different semantics, for example, performance in with argument types musician and song versus performance in with types athlete and competition. Our Approach. We start with a large collection of relational triples, obtained by shallow informa- tion extraction. Specifically, we use the collection of Faruqui and Kumar (2015), obtained by com- bining the OLLIE tool with Google Translate and projecting multilingual sentences back …

Content truncated to 500 words