Finding Answers to Definition Questions

Current researches on Question Answering concern more complex questions than factoid ones. Since definition questions are investigated by many researches, how to acquire accurate answers still becomes a core problem for definition QA. Although some systems use web knowledge bases to improve answer acquisition, we propose an approach that leverage them in an effective way. After summarizing definitions from web knowledge bases and merge them to a definition set, a two-stage retrieval model based on Probabilistic Latent Semantic Analysis is produced to seek documents and sentences in which the topic is similar to those in definition set. Then, an answer ranking model is employed to select both statistically and semantically similar sentences between sentences retrieved and sentences in definition set. Finally, sentences are ranked as answer candidates according to their scores. Experiments indicate following conclusions: 1) specific summarization technologies improves definition QA systems to a better performance; 2) topic based models can be more helpful than centroid-based models for definition QA systems in solving synonym and data sparse problems; 3) shallow semantic analysis is effective to find discriminative characteristics of definitions automatically.