Keyword question answering system with report generation for linked data

摘要

We introduce a question-answering system that responds to a keywords-query by extracting information from linked data and generating reports in natural language (NL). Using entity disambiguation and distributed word similarity, we matched each keyword to a related entity and property in linked data. To extract keyword-related information, we used the entity and property to generate a SPARQL query. NL generation was performed using a NL generation template database. In our experiment, our system answered 95.1% of the user keyword questions in reasonable NL report.