Grammatical Framework (GF) for MT in Sublanguage Domains

Grammatical Framework (GF) for MT in sublanguage domains Janna Khegai Department of Computer Science, Chalmers University of Technology, SE-41296, Gothenburg, Sweden janna@cs.chalmers.se Abstract Grammatical Framework (GF) is a meta-language for multilingual linguis- tic descriptions, which can be used to build rule-based interlingua MT ap- plications in natural sublanguage domains. The GF open-source package contains linguistic and computational resources to facilitate language en- gineering including: a resource grammar library for ten languages, a user interface for multilingual authoring and a grammar development environ- ment. 1 Introduction Grammatical Framework (GF) is a grammar formalism that can be used to build rule- based interlingua MT applications in natu- ral sublanguage domains (Burke & Johan- nisson, 2005; Bringert, Cooper, Ljungl¨of, & Ranta, 2005; Caprotti, 2006). The GF im- plementation takes functional programming approach using a semantic model that could be described in Type Theory (Ranta, 2004). GF provides a powerful meta-language suit- able for describing both natural and formal languages (Ljungl¨of, 2004). The core of a GF grammar is a language- independent interlingua, called abstract syntax. It models a domain by declar- ing categories and relations over them using the notation of functional programming lan- guages. Abstract syntax is the most crucial and difficult part of grammar writing. An- other part, called concrete syntax, maps abstract syntax into strings of natural lan- guage. Every language has its own defini- tion for the given function, called lineariza- tion. Values returned by linearization could be not only strings, but also records and tables, see section 5.2 for some examples. This is especially important for expressing such language-specific features like inflec- tions, morphological parameters and discon- tinuous constituents without affecting the abstract part. Speaking of language-specific lower-level details we want to point out that it would be unreasonably tedious to descend to such details every time we write a GF gram- mar. To address the problem a standard li- brary for the GF language, called resource grammar library, is provided. It decreases grammar development cost by code reuse, guaranteed grammaticality and raising the abstraction level of the task. Resource grammars are now implemented for ten lan- guages: Danish, English, Finnish, French, German, Italian, Norwegian, Russian, Span- ish and Swedish, see Fig. 1. They have been developed in parallel and share the same interface for common rules and cate- gories, which makes implementation of both resource and application grammars easier (Ranta, to appear, 2005). A resource grammar describes a language in general: the basic morphological and syntactical rules applicable to any domain. An application grammar, on the other hand, describes a particular sublanguage do- main, for example, a set of math problems (Caprotti, 2006) or a local transport net- work (Bringert et al., 2005). To write an ap- Figure 1: GF performs MT of interlingua type. Ten languages have general-purpose resource grammars that conform to a common interface. plication grammar using the resource library one need to be a domain-expert, but from the linguistics point of view, it is enough to be a fluent speaker of a language. …

Content truncated to 500 words