Get-In: An Integrated-Ontology-Based Approach for Patent Search
Patents have become increasingly critical to success of many businesses because they create competitive edge
from their exclusive rights granted for inventions – i.e., new technical solutions to problems. As they offer knowledge-rich
data, many researches has been extensively conducted to classify and search for their technical information. However, the
relationships between patent documents and classification schemes have inadequately exploited. Even well-known patent
search systems, such as USPTO, only search for technical information by exploiting either classification scheme or patent
document – without integrating them together. This explains why user queries have been insufficiently analyzed to improve
the accuracy of search results. In this paper, we address such limitations. Our contribution is threefold. First, we propose the
GeT-In ontology that consists of four concepts: Class, Document, Phrase, and Term, to link semantic relationships between the
classification scheme and the patent documents. Second, we propose a method to construct our ontology from a well-known
classification scheme (USPC). Our method does not perform rule-based matching, thus reducing much manual effort
compared to traditional ontology-construction approaches. Third, we propose our patent-search method using our GeT-In
ontology. Our experiments with 500 patents showed that the result yields up to 8% increase in terms of F-measure when
compared to traditional keyword-based approaches.
Keywords— Integrated-Ontology-based Patent Search, Ontology Construction, Information Retrieval.