Passage retrieval for question answering using sliding windows
Mahboob Alam Khalid, Suzan Verberne · 2008
The information retrieval (IR) community has investigated many different techniques to retrieve passages from large collections of documents for question answering (QA). In this paper, we specifically examine and quantitatively compare the impact of passage retrieval for QA using sliding windows and disjoint windows. We consider two different data sets, the TREC 2002--2003 QA data set, and 93 why-questions against INEX Wikipedia. We discovered that, compared to disjoint windows, using sliding windows results in improved performance of TREC-QA in terms of TDRR, and in improved performance of why-QA in terms of [email protected] and MRR.