Retrieving and Exploiting Lexical Semantics for a Comprehensive Corpus of Mathematical Key{phrases: a Programmatic Paper Preliminary Version

In this paper we describe linguistic models and methods, which are being developed for semi{ automatic exploitation and enriching of a large and comprehensive (existing) corpus of English mathematical key{phrases. The starting point is the need for enhanced comparison of key{phrases, meaning, among others, abstraction over intersectivity and movement of modiiers, resolution of prepositional phrase attachment (PPA) ambiguity and indeterminateness of the relations within productive compounds, and isolation of sub{phrases. All this implies the presence of a substantial amount of lexical semantic information (LSI) or LSI tagging. However, no large store of LSI is a priori available. We aim at exploiting the dependencies between the lexical characteristics of the corpus, and the conceptual structure of the domain which is described by the corpus (i.e. mathematics), in order to semi{automatically derive a lexicon in which the lexical entities are modelled to describe the conceptual structure of the domain. We expect such a lexicon to be useful for other types of mathematical texts besides the key{phrases. Important issues in our reasoning will be the kind of LSI that is needed, and creating a model in which limited hand{ tagging forms the base of automatic retrieval of LSI.