What Works Better for Question Answering: Stemming or Morphological Query Expansion?
Matthew W. Bilotti, Boris Katz, Jimmy Lin · 2004
How do dieren t information retrieval techniques aect the performance of document retrieval in the context of question answering? An exploration of this question is our overall research goal. In this paper, we specically examine strategies for coping with morphological variation. This work quantitatively compares two dieren t approaches to handling term variation: applying a stemming algorithm at indexing time, and performing morphological query expansion at retrieval time. We discovered that, compared to the no-stemming baseline, stemming results in lower recall, and morphological expansion yields higher recall. By separately weighting dieren t term variants, we were able to achieve even higher recall, which opens the door to interesting question analysis algorithms for sophisticated query generation. Another signican t contribution of our work is the development of a reusable question answering test collection to support our experiments.