Differentiating Types of Verb Particle Constructions

1 Differentiating Types of Verb Particle Constructions Jon Patrick and Jeremy Fletcher Sydney Language Technology Research Group School of Information Technologies University of Sydney Sydney, Australia, 2006 jonpat@it.usyd.edu.au, jfletch1@it.usyd.edu.au Abstract A verb particle construction (VPC) classification scheme gleaned from linguistic sources has been used to assess its usefulness for identifying issues in decomposability. Linguistic sources have also been used to inform the features suitable for use in building an automatic classifier for the scheme with a series of good performance results. The notions of how to define the task of computing phrasal verbs are discussed and new proposals are presented. 1 Introduction Our area of research focuses on verb-particle constructions (VPCs), a sub-section of multi-word expressions (MWEs). MWEs are a generic term for the group of expressions that include idioms (e.g. over the moon), lexically fixed expressions (e.g. ad hoc), light verb constructions (e.g. make a mistake), institutionalised phrases (e.g. kindle excitement), and verb-particle constructions (e.g. run away). All these expressions have in common the occurrence of words adjacent to each other that would be more frequent than if they were simply random words put together. Hence the words which constitute them have some particular meaning together that they would not have apart. VPCs consist of a simplex (single-word) verb, and a particle, whether preposition or adverb. A particular subset of interest are “phrasal” verbs which are considered to be non-decomposable structures where the meaning is in the whole and not the parts of the phrase (Dixon, 1978). Previous research into the area of MWEs, shows a number of places in which there has been a lack of research. The area which is of interest to us is the approach of bringing reliable resources and specific encodeable linguistic knowledge for use in feature selection in supervised learning tasks of categorisation and WSD. 2 Previous Research There has been only limited research done in the field of MWEs in computational linguistics, mostly focusing on VPCs; although much research (Abeillé, 1988, 1995; Barkema, 1994; Wehrli, 1998) has been done on such MWEs as idioms (Abeillé, 1995; Barkema, 1994; Wehrli, 1998) and light verbs (Abeillé, 1988) within the traditional linguistics field. MWEs are very idiosyncratic constructs making progress in their computation difficult. They have been called “a serious problem for many areas of language technology” (Copestake, Lambeau, Villavicencio, Bond, Baldwin, Sag, Flickinger, 2002) “unpredictable” (Baldwin & Villavicencio, 2002) and “a pain in the neck” (Sag, Baldwin, Bond, Copestake and Flickinger. 2002), and thus many applications of computational linguistics such as machine translation put multiword expressions in the “too- hard” basket. However, the use of MWEs in natural language is widespread, and thus comes about the need for what has been called “a robust, structured handling [of MWEs]” (Baldwin & Villavicencio, 2002, Calzolari, Fillmore, Grishman, Ide, Lencu, MacLeod, and Zampolli. 2002). Much of the previous work on VPCs has revolved around tasks besides WSD, focused on the extraction of multiple word constructs from corpora (Abeillé,1988, Baldwin & Villavicencio, 2002, Maynard & Ananiadou, 1999.) which …

Content truncated to 500 words