A Meta-grammar for CCG

A Meta-grammar for CCG Mark FOREMAN Dept. of Electrical and Electronic Eng. The University of Adelaide Adelaide SA 5005 Mark.Foreman@csiro.au Daniel McMICHAEL Information Enhancement Group CSIRO ICT Centre Waite Road, Urrbrae SA 5064 Daniel.McMichael@csiro.au Abstract Applying CCG to domains outside of linguistics could require different sets of combinators to be developed for each domain. The meta-grammar described in this paper aims to assist such development by enabling simple, succinct expression of both existing and new combinator definitions. It favours the development of an easily-configurable, one- time-coded module that can perform CCG combinations for any combinator set of the researcher’s choosing. A preliminary implementation shows both the feasibility and potential of the meta-grammar. 1 Introduction The merits of Combinatory Categorial Grammar (CCG) have been established via natural language parser implementations like that of Hockenmaier and Steedman (2002) and Clark and Curran (2004). But recent findings show that categorial grammars based on CCG also display promise in domains outside of linguistics (McMichael et al., 2004). Over the years, new combinators have been developed to extend the system of pure categorial grammar (Steedman 2003), but although the set of combinators for CCG seems to have stabilised, this same set may not necessarily be applicable to analyses in other domains. In fact, McMichael et al. introduce two combinators – functional application and modificational application, both defined in section 2.1 – that are not part of the existing set of CCG combinators. Additionally, functional application cannot even be cleanly defined via a traditional combinator pattern. It is partly this inability of the existing techniques to cleanly define new combinators that motivates the proposal of a meta-grammar for combinator specification. In this paper, we explain the motivations for and give a specification of the meta-grammar, along with complete examples of how it applies to new and existing combinators. Lastly, we examine the workings and potential of a preliminary implementation. The remainder of this section, however, presents a brief introduction to CCG. 1.1 A Practical Introduction to CCG CCG operates by first assigning a syntactic category to each word in the sentence, as will be demonstrated in the proceeding example borrowed from Hockenmaier (2003). At this point, the notation for describing categories should be observed. Assuming the simplistic subject-verb-object (SVO) pattern for English, the phrase “buys shares” will form a complete sentence S if it is preceded by a noun phrase, and we write this as: buys shares ├ S\NP So the phrase “buys shares” can be thought of as a function that takes a noun phrase NP as an argument to its left and returns a sentence S. Furthermore, “buys” will form a S\NP if it is followed by a noun phrase, and this is denoted as: buys ├ (S\NP)/NP In doing this, we have eliminated the need for a separate verb category V, leaving us with the following category assignments: John ├ NP buys ├ (S\NP)/NP shares ├ NP Formally, a category may be either atomic (S, NP, etc) or complex (S\S, (S\NP)/NP), etc). Complex …

Content truncated to 500 words