Research And Application Of Topic Mining Technology Based On Patent

In recent years, with the rapid development of information technology and scientific research, the application and grant of patent has skyrocketed, public known patent information data appeared explosively, nowadays, people have realized the patent information is one of the important source of information. With a mass of patent be invented, an effective and accurate method to extract the useful knowledge from patent information is needed, as same as a more reliable method to analyze or inference extractive specific information, finding out the valuable information of the patent and mining depth and effective of patent data, has gradually became a research hotspot.This paper is based on patent data, research on how to mine data specially from patent and inventor by using the relevant knowledge of information retrieval, including discover the influential inventors, patent research hotspot, or the interests of inventors, and recommend inventors to others based on interest of inventor. First of all, this paper studies and analyzes the correlation of homogeneous network Page Rank algorithms based on relevance, then we improve this method and propose a Inventor-Ranking algorithms to rank inventors which is applicable to patent data, this algorithm can not only makes full use of the relationship between the Inventors, but also the relationship of the inventor and the patent, and it is more efficient, the sort result is more conforms to reality. Secondly, we makes full use of the structure characteristics of patent information, propose a algorithm to discover the patent technology hotspot by using the probability distribution based on the LDA topic model, in the actual application scenarios, we find that the algorithms using LDA topic model to detect hot keywords is more accurate and comprehensive. Finally, we propose a inventor personalized recommendation algorithm based on topic model, establish topic model to represent the inventors, calculate similarity of the inventors’ probability distribution and then sort inventors, find out the recommended list of inventors, We make a experiment and compare it to vector space model(VSM) and hidden markov model(HMM), the result shows that, the inventor personalized recommendation algorithm based on LDA topic model has a higher accuracy, a better performance, and it can meet inventors’ personalized requirements better.