SALS-SIG Research Seminar

The Concept of Linguistic Blackbox Revisited

Abstract:

Text generation systems are often structured around two major components: a content planner and a linguistic realizer. This has led many researchers in the field to refer to the "linguistic blackbox": a (preferably) portable module for linguistic realization that is defined strictly in terms of input and output. In this presentation, I will try to shed new light on the concept of linguistic blackbox based on two main observations. Firstly, the blackbox ought to be described not only in terms of input and output, but also in terms of its "paraphrastic capabilities." Secondly, designing a linguistic blackbox as a paraphrase engine has far-reaching consequences for the design of the overall text generation system. In support of this, I will briefly introduce a paraphrase-based lexicalization model for text generation (the inside of the blackbox), showing how such model can free the content planning component from tasks it is commonly expected to perform.

Bio:

Alain Polguère did his studies in formal and computational linguistics in France and Canada. He completed his PhD dissertation in 1990 at the University of Montreal. In parallel with his studies, he also worked in R&D for several years in Montreal, on the development of multilingual text generation systems. Then, from 1991 to 1995, he taught computational linguistics and lexicology at the National University of Singapore. Since 1995, he has been teaching in the Departement of Linguistics and Translation at the University of Montreal. Alain Polguère's current research focus is on both text generation and computational lexicography. He is a member of the research group "Observatoire de Linguistique Sens-Texte" ('Observatory of Meaning-Text Linguistics').

Enquiries: sals@mri.mq.edu.au