Invited Talk: Lexicon Embedded Syntax

This paper explores the notion of lexicon embedded syntax: syntactic structures that are preassembled in natural language lexicons. Section 1 proposes a lexicological perspective on (dependency) syntax: first, it deals with the well-known problem of lexicon-grammar dichotomy, then introduces the notion of lexicon embedded syntax and, finally, presents the lexical models this discussion is based on: lexical systems, as implemented in the English and French Lexical Networks. Two cases of lexicon embedded syntax are then treated: the syntax of idioms, section 2, and the syntax of collocations, section 3. Section 4 concludes on the possible exploitation of syntactic structures that can be extracted from lexical systems. 1 Lexicological Perspective on Syntax 1.1 Lexicon-Grammar Dichotomy The task of modeling languages is often equated with a task of writing so-called grammars. This is clearly demonstrated by the fact that most theoretical proposals in modern linguistics are designated as specific types of grammars: Generative Grammar, Case Grammar, Lexical Functional Grammar, Word Grammar, Generalized Phrase Structure Grammar, Construction Grammar(s), Role and Reference Grammar, Functional Discourse Grammar, etc. (Polguère, 2011, pp 82–83). It should be noted that this focalization on an allencompassing notion of grammar runs deep. For instance, the 1795 law that created the school of oriental language studies in France (INALCO1) specified as follows the linguistic descriptive task assigned to its professors: http://www.inalco.fr “Lesdits professeurs composeront en français la grammaire des langues qu’ils enseigneront: ces divers ouvrages seront remis au comité d’instruction publique.”2 No mention of a need to compile dictionaries for oriental languages, as if it were natural to designate with the term grammar the main tool to be used by XVIIIth century officials and merchants for communicating with “locals”. It should be stressed that this rather confusing notion of Grammar – with a capital G – is extremely broad and encompasses the set of all linguistic rules that make up a natural language. It is distinct from the grammar as a language module that stands in opposition with its functional counterpart: the lexicon. Both linguistic modules have been loosely characterized as follows by O. Jespersen – in terms of their corresponding fields of study: “[g]rammar deals with the general facts of language, and lexicology with special facts” (Jespersen, 1924, p 32). In the present discussion, we will strictly abide by the above characterization and consider the grammar of a language as being the system of all general rules of that language – i.e. rules that are not properties assigned to given words – and the lexicon of that language as being the system of all its word-specific rules. It is a well-established fact that there exists a blurry demarcation between grammar and lexicon (Keizer, 2007). Rules that are specific to linguistic entities that present analogies with “words” but are not strictly speaking lexical units are less lexical in nature and possess a certain grammatical flavor. For instance, rules that account for the properties ‘Said professors will elaborate in French the grammar of languages they will be teaching: these various books will be submitted to the public instruction committee.’