Towards Incomplete SPARQL Query in RDF Question Answering - A Semantic Completion Approach

RDF question/answering(Q/A) system allows users to ask questions in natural language on a knowledge base represented by RDF and retrieve answers. A common problem in RDF Q/A is that existing works tend to translate a natural language question into an incomplete SPARQL query, which means that SPARQL queries may not fully understand user's ideas. For example, some triple patterns may be missing in the question translation stage. In this poster, we first present a siamese adaptation of the Long Short-Term Memory(LSTM) network to detect whether the SPARQL query generated by the RDF Q/A system is complete. Then, for incomplete queries, we propose a Markov-based method to supplement SPARQL queries. Finally, we compare our approach with some state-of-the-art RDF Q/A systems in the benchmark dataset. Extensive experiments confirm that our method improves the precision significantly.