Research On Cross Language Patent Text Analysis

Large amount of patents are a huge information source, they are very important in fields of technology, economy and law. With the tendency of the information and technology globalization, foreign patents sharing and utilization, foreign patents infringement become more and more concerned, the language problems and huge number of patents become the most insuperable barrier for using foreign patents. Cross language patent text analysis can solve the problem of searching related foreign patents accurately and can analyze foreign patent information effectively, but in China, there hasn't any over all work on cross language patent text analyzing and related realization.This paper focused on the cross language patent analysis, the related detail skills and offered the implementing examples. The research work was on basis of parallel Chinese and English patent collections.A machine readable cross language dictionary was made to solve the language problems and at the mean time, ambiguity elimination was analyzed based on the cross language patent corpus. Based on the cross language dictionary, the keywords extraction, text representation and text similarity of cross language patents were discussed. The language synonym dictionaries were used to reduce dimensionality and unite the languages.Cross language patent mapping is an effective solution for fetching precise related patents and in the mean time patent clustering is used widely for patent document analysis and patent invalidity retrieval, this paper took the text mapping and clustering as examples to analyze the cross language patents.During the patent mapping, the text similarity was used and the peculiar structure and taxonomy feature of patents was taken into count to make the mapping results more accurately. The taxonomy feature was also used to evaluate the clustering result.In the end, cross language patent analysis system was established and an analysis example was taken. Chinese and English patent collections were extracted as the example source data, the bilingual mapping and clustering results were given and evaluated.