作者: Leila Kosseim , Guy Lapalme , Luc Plamondon
DOI:
关键词:
摘要: This year, we participated to the Question Answering task for second time with QUANTUM system. We entered 2 runs main (one using web, other without) and 1 run list (without web). essentially built on last year’s experience enhance The architecture of is mainly same as year: it uses patterns that rely shallow parsing techniques regular expressions analyze question then select most appropriate extraction function. function applied one-paragraph long passages retrieved by Okapi extract score candidate answers. Among novelties added this year a web module finds exact answers high-precision reformulation anticipate expected context