Research On The Patent Map Based On Domain Ontology

Patent documentation is strategic information with economic and technological values. In accordance with the statistics by World Intellectual Property Organization, over 90% of the information of inventions and creations are reflected through patent documentation, 90%~95% of research achievements are contained in it. Patent map is a vital tool in patent analysis, which can absorb large amounts of patent documentation and present it in a visual way, by means of searching, screening, classifying and analyzing. This can enlighten the technological experts, inspire their innovation and even help to find new technologies.Numerous are the researches on patent maps and relevant software both at home and abroad. Most of the software, when forming the key process of"semantic measurement of patent texts"of patent maps, is based on text similarity algorithm of the vector space model. Given that patent documentation has such features as wide coverage, as well as excessive infrequent and obscure words, utilization of only vector space model for patent text similarity calculation can only achieve limited effect. For this, domain ontology is necessary to be adopted for semantic analysis of patent documentation. Main researches and contribution of this paper are as follows:(1)By exploration and analysis of all kind of methods of ontology construction, this paper follows common principles of it, induces step of it, and by borrowing ideas from the existing structures of WordNet and making use of existing tools for ontology construction, constructs domain ontology in biological medicine.(2)The paper puts forward text similarity algorithm based on domain ontology. It uses revised word disambiguation algorithm of Lesk, chooses proper meanings, and then calculate semantic similarity by using hierarchical relationship of words meaning in domain ontology. Meanwhile, based on this, the paper utilizes Hungarian algorithm of similar matrix to compute text similarity. Experiments show that this method can obviously improve computing accuracy.(3)On the basis of above content, this paper discusses visualization technology of patent maps, puts forward patent maps with scattered structure which uses patent similarity and IPC class number as X- and Y-coordinate; patent maps with technology effect are made on the basis of word meanings and term frequency, and visualization of patent maps is realized.(4)Finally, the paper applies the technological scheme in patent documentation analysis of gene vaccines, and conducts empirical study on practicability and scientificity of patent maps.