The Research On Distributed Multi Agent System For Extracting Patent Information Based On Two Layer Data Source

In order to improve early warning capabilities and high-tech industrialcompetency, we need to build a platform of intellectual property early-warning.Enterprise, government and agency can get technical support with this platform. Tobulid this platform, we need a large amount of patent information. There are somekind of shortcomings in efficiency, the relationship of data source and usability of thecurrent patent extraction system.DII(Derwent Innovation Index) is a summary database based on many databases,the information of DII is integrated by professional patent analysts, has heig-value forpatent analyzing. But for that DII just contain the subject of patent, we can’t getenough information with when research on detail information, so we need todownload the Meta-information from the detail database.In this paper, the patent information extracting based on two-tier databased hasbeen studied with the distributed systems. In the paper, we solve the problems ofcommunicate of distributed system, load balancing and heterogeneous database.Multi-Agent, XML, distributed system techniques are used in this paper. Meanwhile,to improve the ease of use, we bulid a fuction module to generate the XSLT by users’self-labeling. And the the system will generate the statistics files for patent analyzing.The major contributions of the thesis are as follows: First, the system realize thepatent information acquisition from the heterogeneous patent database by kinds oftechnology. Secondly, the research is based on distribute system with high-level ofparallelism, based on the characteristic of patent extracting, the paper constructalgorithm for load balancing. Thirdly, by using DOM, XSLT, anchor and othertechniques, and with the help of the pruning algorithms developed by myself, thesystem can get the template metadata from a deep webpage, then export the XSTL tothe users.