Research On Patent Quality Evaluation Based On Deep Learning

Patent quality evaluation is of great significance to the promotion of patents,and there are two main routes contained in the patent quality evaluation research,one is the study of patent attributes and the other is the study of patent text.At present,there are many theoretical achievements in patent research on patent attributes,but few relatively patent research on patent text.Therefore,this paper mainly studies the patent quality evaluation from the perspective of patent text.the contents of this paper are as follows:(1)A patent keyword extraction algorithm based on improved attention mechanism and patent keywords of the pointer-generator model is proposed.The traditional attention mechanism has the problem of paying too much attention to a few words in the source text,which leads to the problem of fewer and less comprehensive keywords extracted.In order to solve this problem,this paper proposes a patent keyword extraction method based on improved attention mechanism and pointergenerator neural network.This method reduces the weight of words that have been concerned,which can reduce the probability of the model repeatedly paying attention to the same word,achieving a more comprehensive effect of extracted keywords.The data sets in this paper are based on USPTO patent data in the United States,and the results show that the algorithm has high accuracy and low repetition rate.(2)An improved patent knowledge graph construction method based on deep learning is proposed.In the process of extracting patent keywords from the improved pointer-generator model,according to the times of each word that improved attention mechanism pay to,a knowledge graph construction method based on attention intensity and patent keywords is proposed to solve the problem of the mismatch between patents with common keyword attention intensity,so that the contact intensity value between patents is more accurate and achieves better results.The constructed patent knowledge graph can more accurately express the strength of the connection between patents.Based on improved patent knowledge graph,a patent quality evaluation method is proposed.This method calculates the patent contact strength based on the patent knowledge graph,and uses the patent contact strength to improve the traditional patent attribute quality evaluation method.Experiments show that the improved evaluation method has patents that can identify similar to or inherited core patents.