Study On The Method Of Patent Recommendation Based On Automatic Tagging Function Information

Since the patent text contains composition of product structure, design goals, function principle component and other related content, it has always been seen as an important source of knowledge and innovation of product design.Patent is inside the solution domain of practical and innovative which contains Abundant innovation technology of information.The innovation of the design principle and function is the foundation of product innovation. However, in the actual practice of production activities innovatives, due to subject to their own inertia thinking, scope of knowledge and education level limit, innovative designers often can’t breakthrough the traditional mode of thinking of the limit, and unable to determine the solution of ultimate goal to be achieved, unable to find a way to simplify the Staff workload and Inspire designers breakthrough thinking limit. Therefore, how to recommend related patent information of the patent goals and patent realization principle has increasingly become a hotspot. However, because of the natural language optional description do not have a fixed format nested, the description of patent goals and patent realization principle increases the challenge and difficulty.This paper through a large number of patent literature readings, analyzes the Chinese patent structure and format, combined with sentence analysis, lexical analysis, syntactic analysis and related rules and algorithm. This paper uses the research of the Chinese Academy of Science’s ICTCLAS as lexical analysis component to word segmentation in Chinese text, parser components by Stanford University R & D as the syntactic analysis components to get syntax statement. Through the association of scientific effect knowledge and patent text physical quantity, the corresponding relationship between the physical quantity, effect and function has been found, and then made the physical domain physical quantity effect function table. By using the table and combined with the corresponding physical quantity effect algorithm and effect corresponding function algorithm, Correspondence and label the patent function principle information, and finally makes this information available to the patent goals and patent realization principle recommendation. This paper adopts the "Verb +Noun/ Phrase" two-tuples form and "Verb + Noun/ Phrase + the + Noun/Phrase" three-tuples form to express the information. Experiments show that, this method is correct and effective.