Automatic Interpretation of Noun Compounds Using WordNet Similarity

Automatic Interpretation of Noun Compounds Using WordNet Similarity Su Nam Kim1,2 and Timothy Baldwin2,3 1 Computer Science, University of Illinois, Chicago, IL 60607 USA sunamkim@gmail.com 2 Computer Science and Software Engineering, University of Melbourne, Victoria 3010 Australia 3 NICTA Victoria Lab, University of Melbourne, Victoria 3010 Australia tim@csse.unimelb.edu.au Abstract. The paper introduces a method for interpreting novel noun compounds with semantic relations. The method is built around word similarity with pre- tagged noun compounds, based on WordNet::Similarity. Over 1,088 training instances and 1,081 test instances from the Wall Street Journal in the Penn Treebank, the proposed method was able to correctly classify 53.3% of the test noun compounds. We also investigated the relative contribution of the modi- fier and the head noun in noun compounds of different semantic types. 1 Introduction A noun compound (NC) is an ¯N made up of two or more nouns, such as golf club or paper submission; we will refer to the rightmost noun as the head noun and the re- mainder of nouns in the NC as modifiers. The interpretation of noun compounds is a well-researched area in natural language processing, and has been applied in applica- tions such as question answering and machine translation [1,2,3]. Three basic properties make the interpretation of NCs difficult [4]: (1) the compounding process is extremely productive; (2) the semantic relationship between head noun and modifier in the noun compounds is implicit; and (3) the interpretation can be influenced by contextual and pragmatic factors. In this paper, we are interested in recognizing the semantic relationship between the head noun and modifier(s) of noun compounds. We introduce a method based on word similarity between the component nouns in an unseen test instance NC and annotated training instance NCs. Due to its simplicity, our method is able to interpret NCs with significantly reduced cost. We also investigate the relative contribution of the head noun and modifier in determining the semantic relation. For the purposes of this paper, we focus exclusively on binary NCs, that is NCs made up of two nouns. This is partly an empirical decision, in that the majority of NCs occurring in unrestricted text are binary,1 and also partly due to there being ex- isting methods for disambiguating the syntactic structure of higher-arity NCs, effec- tively decomposing them into multiple binary NCs [3]. Note also that in this paper, we 1 We estimate that 88.4% of NCs in the Wall Street Journal section of the Penn Treebank and 90.6% of NCs in the British National Corpus are binary. R. Dale et al. (Eds.): IJCNLP 2005, LNAI 3651, pp. 945–956, 2005. c⃝ Springer-Verlag Berlin Heidelberg 2005 946 S.N. Kim and T. Baldwin distinguish semantic relations from semantic roles. The semantic relation in an NC is the underlying relation between the head noun and its modifier, whereas its semantic role is an indication of its relation to the governing verb and other constituents in the sentence context. There is a significant body of closely-related research on interpreting semantic rela- …

Content truncated to 500 words