Semantic Equivalent Relation Phrase Extraction for RDF Predicates
Lei Xu, Yunpeng Liu, Peng Nie · 2019
Relation phrase mapping can fill the gap between the relation phrase and the RDF predicates. It can be widely used in question/answer system to help the general users ask questions in natural language over an RDF knowledge base. Relation phrase mapping is one of the most important steps to mine the semantic meaning of natural language questions. Instance extraction has been well studied by some literal work whereas the semantic meaning for relation phrase has the same importance. The gAnswer system has conducted a paraphrase dictionary to fill the gap between the relation phrase and the RDF predicates. However, the precision of the paraphrase dictionary is unsatisfactory for natural language question answer. In this paper, we adopt the structure of gAnswer and present a method to build a more accurate paraphrase dictionary using vector space model. The result confirms that our method can generate a large number of relation phrase and improve the precision at the same time.