Automatic Collection of Natural Language Expressions of Relations for Natural Language Interface
Yong-Jin Han, Se‐Young Park, Seong-Bae Park · Jeongbo gwahaghoe nonmunji. so'peuteuweeo mich eung'yong · 2011
One of the critical problems in natural language interfaces (NLI) to structured data is to handle the large variability in the way that a relation can be expressed. Existing approaches to solve this problem rely on manual process to adapt NLI to a new domain. The manual approaches can be efficiently applied to small structured data. However, nowadays, large volumes of graph-structured data such as RDF and OWL have been produced from a variety of areas on the web. It is not easy to adapt NLI to such large graph structured data. As a solution of the variability problem in NLI, this paper proposes an automatic method to identify relations in natural language queries for NLI to graph structured data. A relation between two entities is uniquely defined as an edge linking the entities on graph-structured data. The proposed method automatically discovers natural language expressions for given relations by querying entity pairs linked with the relations from a corpus. A relation corresponding to an edge is modeled as a collection of frequently appeared expressions. In the experiment, automatically collected expressions are applied to a NLI and the NLI showed comparable results with manually approaches.