AUTOMATIC EXTRACTION OF SEMANTIC RELATIONS FROM PATENT CLAIMS
Shih-Yao Yang, Szu‐Yin Lin, Shih-Neng Lin, Chi-Feng Lee, Shian-Luen Cheng, Von‐Wun Soo · Int. J. Electron. Bus. Manag. · 2008
In this paper, we propose a set of techniques to extract the semantic relations from patent claims and extend it to facilitate the comparisons among patent claims on the basis of the semantic structures. The system could parse and annotate every claim sentence and extract the desired semantic information from the patent claim texts using the regular expression pattern matching techniques. The extracted semantic information is translated into a semantic structure in terms machine readable format OWL for knowledge sharing. A total 48 regular expressions in eight kinds of domain specific regular expressions are identified and are implemented for testing. The evaluation experiments conducted on the chemical mechanical polishing domain resulted in the performance of 69% recall and 86% precision on average. The system spends on average 0.97 seconds to complete the information extraction from the claim in each patent document.