Computational Linguistics Colloquium
Thursday, 3 February 2011, 16:15
Modelling linguistic complexity
Laboratoire Parole et Langage, CNRS & Université de Provence
Evaluating the complexity of a language or, for a given language, the complexity of a specific construction remains a problem. Linguists address this issue in different perspectives: typology (Are all languages equally complex?), description (What constructions are complex? How to characterize the complexity of a subdomain such as syntax, phonology, morphology, etc. ?), or operational level (Is language complexity a function of subdomains complexity?).
On their side, psycholinguists also study this question from the processing point of view (How difficult to process is a given construction ? How complex is language acquisition?, etc.). And more recently, computational linguists also investigated this question: some complexity measurements have been proposed, evaluating the depth of the syntactic structure, the complexity of the parser, etc.
Each perspective bears an element of answer of the problem, and it is necessary to propose a general framework bringing together these approaches into a unique model. Doing this is interesting for two reasons: theoretically (such a framework explains complexity in a multi-parametric perspective) and computationally (it becomes possible to quantify linguistic complexity).
We present in this talk, after a general overview of linguistic complexity, an approach focusing on syntactic complexity and integrating the different approaches. It consists in giving the different complexity parameters a homogenous representation, starting from which a computational model can be built.
If you would like to meet with the speaker, please contact