Similar Patent Search Method Based On A Functional Information Fusion

In the Internet era,patent data is growing at an alarming rate.How to obtain valuable technical information from massive patent data and find similar patents has become a hot topic at present.The most fundamental innovation of products is the innovation of functional principle.Learning from the functional information of existing patents plays an important role in the design of product innovation and patent circumvention.In recent years,the methods proposed by domestic and foreign research scholars for obtaining patent information either cannot be popularized,or the characteristics of patent texts are not deeply analyzed;similar patent searches are proposed either starting from patent citations or calculating the contents of patent texts,but not highlight the importance of the functional information.In response to the above problems,this paper proposes a similar patent search method that fuses functional information.This paper uses the patent abstract and the content of the invention as the original data source,through the analysis of the expression of functional information,the phrase "verb + noun / phrase" is used to describe the functional information,ie "operation + object".This paper first divides paragraphs into sentences,then obtains long sentences containing function information,short phrases containing functional information and functional phrases in order.Finally,a similar patent search method based on functional information is proposed.The main innovations are as follows:(1)by analyzing the characteristics of mutual information technology and patent text,a candidate word extraction method based on mutual information is proposed,which is a word that makes a larger contribution to judging whether a sentence contains functional information.(2)the characteristics of the original corpus are deeply analyzed.It is proposed that the problem of judging whether the sentence contains functional information is converted to the problem of short text classification,and a short text classification method based on the candidate words is proposed.(3)combining the commonly used verbs,nouns which are commonly used to express functional information and Stanford Parser,the functional phrases extraction rules and functional phrase extraction rules are summarized.(4)uses word2 vec to calculate the similarity between words,uses the results of dependency-syntax analysis to pick words for similarity calculation.The word similarity between specific dependency pairs accounts for a greater weight in the final calculation of sentence similarity to improve the impact of patent function information.Experiments show that the short text classification method based on candidate words proposed in this paper can find out most of the functional information sentences.The extraction method of functional information phrases can also extract most functional phrases.The similar patent retrieval method of fusion functional information proposed in this paper improves the basis calculating the similarity of the entire document.,enlarges the function of the functional information,and uses the functional information as the main basis to search for similar patents.