Collocations in Multilingual Natural Language Generation: Lexical Functions meet Lexical Functional Grammar

Collocations in Multilingual Natural Language Generation: Lexical Functions meet Lexical Functional Grammar Franc¸ois Lareau Mark Dras Benjamin B¨orschinger Robert Dale Centre for Language Technology Macquarie University Sydney, Australia Francois.Lareau|Mark.Dras|Benjamin.Borschinger|Robert.Dale@mq.edu.au Abstract In a collocation, the choice of one lexical item depends on the choice made for another. This poses a problem for simple approaches to lex- icalisation in natural language generation sys- tems. In the Meaning-Text framework, recur- rent patterns of collocations have been char- acterised by lexical functions, which offer an elegant way of describing these relationships. Previous work has shown that using lexical functions in the context of multilingual natural language generation allows for a more efficient development of linguistic resources. We pro- pose a way to encode lexical functions in the Lexical Functional Grammar framework. 1 Introduction Natural Language Generation (NLG) is the gener- ation of natural language text from some underly- ing representation: numerical data, knowledge bases, predicate logic, and so on. Multilingual Natural Language Generation (MNLG) is NLG where the output is in more than one language. Most NLG systems are in some way modular (see Reiter and Dale (2000) for a discussion of typical architectures); one advantage to modularity is the scope for sepa- rating language-independent components from those which are language-dependent, making it possible to add multilinguality with much less work than would be involved in building a new system from scratch (Bateman et al., 1999). Such claims have been made since the very first MNLG systems; the FoG system generating weather forecasts in English and French (Bourbeau et al., 1990) is a case in point. Consequently, MNLG has been applied for a large number of text types: government statistics reports (Iordanskaja et al., 1992), technical instruction man- uals (Paris et al., 1995), fairy tales (Callaway and Lester, 2002), museum tours (Callaway et al., 2005), medical terminology (Rassinoux et al., 2007), codes of practice (Evans et al., 2008), and so on. Marcu et al. (2000), in reviewing some of the ear- lier work, comment that MNLG systems need to ab- stract as much as possible away from the individual language generated: If an [MNLG] system needs to develop language dependent knowledge bases, and language dependent algorithms for con- tent selection, text planning, and sentence planning, it is difficult to justify its eco- nomic viability. However, if most of these components are language indepen- dent and/or much of the code can be re- used, an [MNLG] system becomes a viable option. Bateman et al. (1999) similarly emphasise the impor- tance of reducing language dependence. One kind of abstraction generalises across language-specific collocations: for example, we might note that heavy rain, strong wind or intense bombardment all refer to the intensification of some phenomenon, as similarly does the French pluie bat- tante (‘beating rain’), but the particular intensifier used is determined by collocational appropriateness. These kinds of collocations are modeled within the Meaning-Text Theory (MTT) framework via lexical functions (LFs) (Mel’ˇcuk, 1995); for some lexeme L, the above semantic notion of intensification or Fran¸cois Lareau, Mark …

Content truncated to 500 words