Research On Acquisition And Application Of Deepknowledge Based On Triz Theory

Designers cannot generate real innovative products with their singleprofessional knowledge since they lack creativity in traditional products designof conceptual phase. Although TRIZ Theory could guide designers to innovate, itcannot be ease to master completely in practice. Patents contained in traditionalpatents database are the main creative resources, but they are hard to search andemploy because the traditional patents database includes countless patentdocuments which are classified as various disciplines. Therefore, it is verysignificant that expanding and updating the knowledge database which isconstructed by extracting, recombining, and analyzing according to TRIZtheory’s knowledge because it is in favor of practical application andself-improvement of TRIZ theory, assist people to master universal law from thepoint of view of innovation theory in innovative designs, and motivate designers’divergent thinking on the basis of analogizing from past successful cases.The deep knowledge acquisition model based on TRIZ theory was proposedin this thesis. The model based on TRIZ theory acquired deep knowledge frompatents resource by means of data mining technology. The deep knowledgeacquiring process contributes to the interdisciplinary application of patentdatabase and the research of knowledge discovery and reusing. The process ofapplying and popularizing TRIZ theory among ordinary designers was alsoexplored.The thesis was discussed according to implementation methods and keytechnologies of each sub-module of the model, containing several parts below:patent text extraction module, text classifier module, and depth-knowledgemining module.Firstly, the patent text extraction module introduced how to acquire patentsdatabase information from SIPO’s web pages, and store the information in the database. Patent text extraction was the precondition of the whole process ofknowledge mining. If the patent abstracts and basic information we needed wereextracted incorrectly, the patent database could not be able to be constructed andanalyzed. The study of the module was crucial for the future research.Secondly, text classifier module mainly realized on patent text classification.This part was mainly talking about the classification methods and process fromtwo aspects, i.e. the artificial and automatic classification. The artificialclassification need people, who mastered some certain TRIZ theory and fieldknowledge, to read patent specification carefully. Abstracts were the targetobjects by computer classification because they could replace basically thewhole patent text’s contents, and the computers’ computation also could begreatly simplified, which was very convenient and practical for the classifier intesting stage. The patents in this module were analyzed and classified accordingto innovation principles.Thirdly, deep knowledge was extracted from classification results in thedepth-knowledge mining module. Mining process was analyzed by a deepknowledge template wizard on the basis of reading patent specification, and thenthe results were stored in the case database.Finally, the software，named DKMining, was constructed. The softwarerealized functions of each module, and was able to search, delete, modify andupdate the information in the patent database and case database. The thesiscombined with a specific patent to verify the theory research above, so thesoftware system had certain feasibility.