Syntactic Analysis of Phrasal Compounds in Corpora: a Challenge for NLP Tools

Syntactic Analysis of Phrasal Compounds in Corpora: a Challenge for NLP Tools Carola Trips Universität Mannheim L13, 9, 68131 Mannheim ctrips@mail.uni-mannheim.de Abstract The paper introduces a “train once, use many” approach for the syntactic analysis of phrasal compounds (PC) of the type XP+N like “Would you like to sit on my knee?” nonsense. PCs are a challenge for NLP tools since they require the identification of a syntactic phrase within a morphological complex. We propose a method which uses a state-of-the-art dependency parser not only to analyse sentences (the environment of PCs) but also to compound the non-head of PCs in a well-defined particular condition which is the analysis of the non-head spanning from the left boundary (mostly marked by a determiner) to the nominal head of the PC. This method contains the following steps: (a) the use an English state-of-the-art dependency parser with data comprising sentences with PCs from the British National Corpus (BNC), (b) the detection of parsing errors of PCs, (c) the separate treatment of the non-head structure using the same model, and (d) the attachment of the non-head to the compound head. The evaluation of the method showed that the accuracy of 76% could be improved by adding a step in the PC compounder module which specified user-defined contexts being sensitive to the part of speech of the non-head parts and by using TreeTagger, in line with our approach. Keywords: Phrasal compounds, state-of-the-art dependency parser, compounder 1. State of the art This paper problematizes the syntactic analysis of phrasal compounds extracted from English corpora and proposes a “train-once, use-many approach” which uses a state-of-the art dependency parser for the analysis of full sentences and as a compounder for the non-head of phrasal compounds. In many languages compounding is a productive process of word formation. This is the case for example in Ger- man and English where different types of compounds exist with Noun+Noun compounds (NNC) being most frequent. In these languages the structure of NNCs is such that the formal and semantic head is right-headed with the deter- mining or modifying non-head preceding it (e.g. German Haustür ‘door of a house’, English house guest). Although both languages exhibit NNCs, unlike English, German does not graphically split the non-head and the head, which re- sults in ambiguities in automatic analysis. Thus, the task of finding and segmenting unknown compounds in text cor- pora is at stake here, and a number of methods (for different purposes) has been proposed. For example, Brown (2002) dealing with parallel English-German texts in the medical domain, proposes a method for splitting compound words into their constituents based on cognate words in the other language of a parallel corpus. Schiller (2005) discusses an experiment where she used a finite-state morphological an- alyzer to segment compounds which is based on weights for compound segments. Daðason & Bjarnadóttir 2014 de- scribe a method where a decompounder which estimates the probability of a constituent in an Icelandic unknown com- pound on the basis of known compounds …

Content truncated to 500 words