Abstract
In the 21st century, intelligient property has become the key factor of competitive strength of the global economy, and patent document can protect the intelligient property in law effectively. Unfortunately, because of the explosion in patent documents and the claim formats are full of variety, the technologies of patent documents in search、analysis、and comparison have faced some serious bottleneck. In my thesis, I propose an approach to extract the semantic structure of claim and compare their difference on the basis of semantic structures. In the aspect of semantic structure extraction, if the user chooses some patents of the same domain for the patent system to analyze, they need to construct the domain thesaurus and ontology first for further parsing. Second, the system will parse and annotate every claim by natural language processing technics and extract the important information from claims by regular expression. At last, this information is translated into semantic structure in machine readable formats, XML and OWL, for speeding the knowledge sharing and knowledge inference and can be displayed in graph. In the aspect of claim comparison, when the user chooses an invention component in the semantic structure of a patent to query, the system will execute the similarity algorithm to see if there are some similar components exist in other patents. The algorithm will take two components's semantic annotation、semanitc structure、 and attributes into consideration to calculate the similarity of the two components. The experimental results show that the approach can effectively parse and extract the semantic content of claims, and assist users to understand the the focal point of claims by GUI environment. On the other hand, the similarity algorithm can find the similar invention components substantially and provides a different way from keyword search to search patent.