Natural Language Questions and Answers for RDF Information Resources

摘要

We propose Questions and Answers (Q&A) method responding natural language questions about RDF information resource. When a natural language question and an RDF graph are given, keywords are extracted from the question using morphological analysis and keywords are converted to key elements referencing lexica describing correspondence relationships between keywords and elements. A question subgraph containing all key elements corresponding to keywords in the question are extracted from the RDF graph and the question subgraph is searched for an answer. We evaluate performance using an RDF information resource that describes our laboratory.