|
In finding potential patents for litigation, the common practice is taking a picture, inspecting, or using reverse engineering to obtain technical information of a product, and then comparing the product with claims of a patent to confirm an act of infringement. Therefore, the cost of infringement verification is quite high. The present study takes a different approach: using patent number marked on the product to obtain technical information of the product, and comparing the product-relating patent with a potential patent in the art by text mining to find the similarity between those two patents. Therefore, a preparation before patent infringement verification and a search for a list of patents which may use during litigation can be done much faster. The present study uses regular expression for grammar parsing to extract elements in independent claim of the potential patent. The extracted elements are used to compare with the contents in different sections of the product-relating patent, which comprise abstract, independent claims, dependent claims, and specification; thereby, the present study shows that contents in different sections of the product-associating patent have different effect to the analysis results. The results and feasibility of the present study is as follows: in terms of analysis quantity, the analysis quantity of claims in potential patents drops from 1590 to 89, which accounts for 5.6% of the original analysis quantity; further, instead of claims, if potential patents are concerned, the analysis quantity of potential patents drops from 716 to 53, which accounts for 7.4% of the original analysis quantity. In a sample of 48 patents, the analysis result indicates the quantity of patents with high relevance, middle relevance, low relevance , and others are 35 (73%), 4 (8%), 7 (15%), and 2 (4%), respectively. The element extraction of claims provided in the present study may be used in patent monitoring, analysis of freedom to operate or patent validation in the future. |