Learning to Distinguish PP Arguments from Adjuncts

Learning to Distinguish PP Arguments from Adjuncts Aline Villavicencio Computer Laboratory, University of Cambridge J.J Thomson Avenue, Cambridge, CB3 OFD, UK Phone: +44-1223-763642 Fax: +44-1223-334678 Aline.Villavicencio@cl.cam.ac.uk Abstract Words differ in the subcategorisation frames in which they occur, and there is a strong cor- relation between the semantic arguments of a given word and its subcategorisation frame, so that all its arguments should be included in its subcategorisation frame. One problem is posed by the ambiguity between locative prepositional phrases as arguments of a verb or adjuncts. As the semantics for the verb is the same in both cases, it is difficult to differentiate them, and to learn the appropriate subcategorisation frame. We propose an approach that uses se- mantically motivated preposition selection and frequency information to determine if a locative PP is an argument or an adjunct. In order to test this approach, we perform an experiment using a computational learning system that re- ceives as input utterances annotated with log- ical forms. The results obtained indicate that the learner successfully distinguishes between arguments (obligatory and optional) and ad- juncts. 1 Introduction Words differ in the subcategorisation frames that realise their semantic arguments, and a given word may have several different subcate- gorisation frames. The subcategorisation frame includes all the complements of a given word. For instance, the sentences: •(1)John ate •(2)John ate the apple represent the intransitive and transitive frames, respectively, and both are valid frames associ- ated with the word eat. Given that the subcat- egorisation frame of a given word should only include a given constituent if it is an argument, one problem is caused by the ambiguous nature of some constituents, that can be either argu- ments or adjuncts. The ability to distinguish between subcate- gorised arguments and non-subcategorised ad- juncts is of great importance for several applica- tions, such as automatic acquisition of subcat- egorisation lexicons from data, and this prob- lem has been widely investigated. For instance, Buchholz (1998) investigates this task using a memory-based learning approach, where the use of syntactic and contextual features results in a 91.6% accuracy in distinguishing arguments from adjuncts. Brent (1994) looks at the prob- lem from a more psychologically oriented per- spective, trying to simulate the environment available to a human language learner, and us- ing binomial error estimation to derive subcat- egorisation frames for verbs, based on imper- fectly reliable local syntactic cues. This tech- nique is able to capture the fact that the rel- ative frequency of a verb-argument sequence is likely to be higher than that of a verb-adjunct sequence. However, the cues used in the sim- ulations are too simple to achieve high accu- racy. Steedman (1994) suggests the use of se- mantic information to deal with this ambigu- ity, given that syntax should be as close as possible to semantics. Then, given that for a particular language there is a strong cor- relation between the subcategorisation frames and predicate-argument structure of a given word, from the predicate-argument structure of a word it …

Content truncated to 500 words