融合语义特征的高校专利质量预测研究

Research on Patent Quality Predication for Universities Based on Integrate Semantic Features

分析国内外高校专利质量评价的理论与实践研究发现:高校专利申请量剧增,但专利产业化率较低;专利质量预测研究较少,且仅停留在理论层面,对高校专利质量预测更少。因此,建立实用的、易操作的专利质量预测模型对于提高专利质量尤为重要。选取说明书摘要等文本型专利指标和权利要求数量等数字型指标,建立高校专利质量预测指标体系,运用词向量转换理论,深入挖掘文本信息,将文本指标转化为可用的数值参数,在此基础上创新地提出专利质量预测模型--Word2Vec-XGB预测模型。构建的融合语义特征专利质量预测模型,其预测平均准确率为90%以上,适合高校专利申请前预评估,实现高质量专利预测,有助于进一步提高专利质量。

Analyzing the theoretical and practical research on patent quality evaluation in universities domestically and internationally,it is found that the number of patent applications in universities has increased dramatically,but the rate of patent industrialization is relatively low.The research on patent quality prediction is relatively small and only stays at the theoretical level,and there are even fewer predictions of patent quality for universities.Therefore,it is especially important to establish practical and easy-to-operate patent quality prediction models to improve patent quality.This paper select textual patent indicators such as abstracts of specifications and numerical indicators such as the number of claims to establish a patent quality prediction index system for universities.Using word vector transformation theory,the model digs into the textual information,and transforms the textual indicators into usable numerical parameters.On this basis,the paper innovatively put forward a patent quality prediction model-Word2Vec-XGB prediction model.The average accuracy of the predictive model is more than 90%,which is suitable for pre-evaluation before patent application in universities,and it can help to further improve the quality of patents by achieving high-quality prediction.