Automatic Question-Answering (QA) is a sub-field of Information Extraction, and traditionally strongly entwined with statistical methods and language models. With a few notable exceptions, most ground-breaking work in this field has so far been focusing on English. We believe, however, that rule based or hybrid methods may go further than statistical methods alone, and that there is a need for language-specific systems in the area. Experimentally, VISL-style grammatical analysis has thus been used for both the identification of question patterns and the mark-up of source data for searching, using Constraint Grammar parsers like PALAVRAS (Portuguese) or DanGram (Danish). Currently, an academic pilote project for Portuguese is under way, while a Danish system is being developed commercially by Aminova Dialogue Systems and GrammarSoft.

Publications:
  • Bick, Eckhard (2003). "A Constraint Grammar Based Question-Answering System for Portuguese". In: Fernando Moura Pires & Salvador (eds.), Progress in Artificial Intelligence (Proceedings of EPIA'2003, Beja, Dec. 2003), pp. 414-418. Springer