NLG vs. Templates -- Five Years Later

Richard Kittredge
University of Montreal and Cogentex

** special days ** Wednesday 26 April at 11am

Abstract

Natural language generation (NLG) using "deep" language modelling offers many benefits to application builders, but at a price which is often underbid by "shallow" approaches using text templates with word string fillers, nowadays marked up to include some linguistic information.  We will update Reiter's '95 comparison of NLG and template approaches to text production, with emphasis on current concerns in outputting controlled language and synthesized speech.

Short resume

Richard Kittredge received his Ph.D. in formal linguistics and mathematical logic from the University of Pennsylvania in 1969. During the 1970s and 1980s he pioneered the development of practical systems for machine translation and text generation at the University of Montreal, where he is currently Professor of Linguistics.  In 1990 he co-founded CoGenTex, a research and development company specializing in NLG.  His research interests include automatic summarization, multi-lingual text generation and the design of controlled languages for industrial applications.

To top

last updated December 01, 2003 10:37 AM
Sandrine.Balbo@cmis.csiro.au