Grounded Semantic Parsing for Complex Knowledge Extraction

Human Language Technologies: The 2015 Annual Conference of the North American Chapter of the ACL, pages 756–766, Denver, Colorado, May 31 – June 5, 2015. c⃝2015 Association for Computational Linguistics Grounded Semantic Parsing for Complex Knowledge Extraction Ankur P. Parikh∗ School of Computer Science Carnegie Mellon University apparikh@cs.cmu.edu Hoifung Poon Kristina Toutanova Microsoft Research Redmond, WA, USA hoifung,kristout@microsoft.com Abstract Recently, there has been increasing interest in learning semantic parsers with indirect super- vision, but existing work focuses almost ex- clusively on question answering. Separately, there have been active pursuits in leveraging databases for distant supervision in informa- tion extraction, yet such methods are often limited to binary relations and none can han- dle nested events. In this paper, we gener- alize distant supervision to complex knowl- edge extraction, by proposing the first ap- proach to learn a semantic parser for extract- ing nested event structures without annotated examples, using only a database of such com- plex events and unannotated text. The key idea is to model the annotations as latent variables, and incorporate a prior that favors semantic parses containing known events. Experiments on the GENIA event extraction dataset show that our approach can learn from and extract complex biological pathway events. More- over, when supplied with just five example words per event type, it becomes competitive even among supervised systems, outperform- ing 19 out of 24 teams that participated in the original shared task. 1 Introduction The goal of semantic parsing is to map text into a complete and detailed meaning representation (Mooney, 2007). Supervised approaches for learn- ing a semantic parser require annotated examples, ∗ This research was conducted during the author’s intern- ship at Microsoft Research. Information Extraction ID TYPE CAUSE THEME TRIGGER T1 PROTEIN - - BCL T2 PROTEIN - - RFLAT T3 PROTEIN - - IL-10 E1 POS-REG T1 E2 stimulates E2 NEG-REG T3 T2 inhibition (NEG-REG,BCL,RFLAT) (NEG-REG,IL-10,RFLAT) Complex Event (POS-REG,BCL,(NEG-REG,IL-10,RFLAT)) GENIA Event Annotation Figure 1: Given sentence “BCL stimulates inhibition of RFLAT by IL-10”, information extraction focuses on classifying simple relations among entities (top), whereas ideally we want to extract the complex event that captures important contextual information (middle), as exempli- fied by the GENIA event annotation (bottom). which are expensive and time-consuming to acquire (Zelle and Mooney, 1993; Zettlemoyer and Collins, 2005; Zettlemoyer and Collins, 2007). As a re- sult, there has been rising interest in learning se- mantic parsers from indirect supervision. Examples include unsupervised approaches that leverage dis- tributional similarity by recursive clustering (Poon and Domingos, 2009; Poon and Domingos, 2010; Titov and Klementiev, 2011), semi-supervised ap- proaches that learn from dialog context (Artzi and Zettlemoyer, 2011), grounded approaches that learn from annotated question-answer pairs (Clarke et al., 2010; Liang et al., 2011) or virtual worlds (Chen and Mooney, 2011; Artzi and Zettlemoyer, 2013). Such progress is exciting, but most applications focus on question answering, where the semantic parser is used to convert natural-language questions into formal queries. In contrast, complex knowl- edge extraction represents a relatively untapped ap- 756 (a) …

Content truncated to 500 words