Pattern Dictionary of English Prepositions

Proceedings of the 52nd Annual Meeting of the Association for Computational Linguistics, pages 1274–1283, Baltimore, Maryland, USA, June 23-25 2014. c⃝2014 Association for Computational Linguistics Pattern Dictionary of English Prepositions Ken Litkowski CL Research 9208 Gue Road Damascus, MD 20872 USA ken@clres.com Abstract We present a new lexical resource for the study of preposition behavior, the Pattern Dictionary of English Prepositions (PDEP). This dictionary, which follows principles laid out in Hanks’ theory of norms and exploita- tions, is linked to 81,509 sentences for 304 prepositions, which have been made available under The Preposition Project (TPP). Nota- bly, 47,285 sentences, initially untagged, provide a representative sample of preposi- tion use, unlike the tagged sentences used in previous studies. Each sentence has been parsed with a dependency parser and our sys- tem has near-instantaneous access to features developed with this parser to explore and an- notate properties of individual senses. The features make extensive use of WordNet. We have extended feature exploration to include lookup of FrameNet lexical units and VerbNet classes for use in characterizing preposition behavior. We have designed our system to allow public access to any of the data available in the system. 1 Introduction Recent studies (Zapirain et al. (2013); Srikumar and Roth (2011)) have shown the value of prepo- sitional phrases in joint modeling with verbs for semantic role labeling. Although recent studies have shown improved preposition disambigua- tion, they have received little systematic treat- ment from a lexicographic perspective. Recently, a new corpus has been made available that prom- ises to be much more representative of preposi- tion behavior. Our initial examination of this corpus has suggested clear indications of senses previously overlooked and reduced prominence for senses thought to constitute a large role in preposition use. In section 2, we describe the interface to the Pattern Dictionary of English Prepositions (PDEP), identifying how we are building upon data developed in The Preposition Project (TPP) and investigating its sense inventory with corpo- ra also made available under TPP. Section 3 de- scribes the procedures for tagging a representa- tive corpus drawn from the British National Cor- pus, including some findings that have emerged in assessing previous studies of preposition dis- ambiguation. Section 4 describes how we are able to investigate the relationship of WordNet, FrameNet, and VerbNet to this effort and how this examination of preposition behavior can be used in working with these resources. Section 5 describes how we can use PDEP for the analysis of semantic role and semantic relation invento- ries. Section 6 describes how we envision further developments of PDEP and how the data are available for further analysis. In section 7, we present our conclusions for PDEP. 2 The Pattern Dictionary of English Prepositions Litkowski and Hargraves (2005) and Litkowski and Hargraves (2006) describe The Preposition Project (TPP) as an attempt to describe preposi- tion behavior using a sense inventory made available for public use from the Oxford Dic- tionary of English (Stevenson and Soanes, 2003) by tagging sentences drawn from …

Content truncated to 500 words