面向TRIZ的专利自动分类研究

发布时间：2018-09-19 20:48

【摘要】：【目的】通过构建个性化分类体系,研究面向TRIZ应用的专利自动分类方法。【方法】基于主题模型,从宏观、中观、微观三个层面构建面向TRIZ个性化分类体系;通过对不同分类特征项与算法进行组合,挑选分类准确率最高的组合构建初始分类器;采用平滑非平衡数据与特征项降维方式对分类器进行优化,完成对专利的自动分类。【结果】实现半自动构建面向TRIZ的个性化分类体系及基于该分类体系的专利自动分类。在中等数据量级场景下(千条),实现专利自动分类,分类效果综合评价指标高达90.2%。【局限】该方法不适用于数据量较小(百条)时的专利分类;在较大数据量(万条)场景下,该方法的有效性尚未得到验证。【结论】对中等规模专利数据,能快速构建面向TRIZ的分类体系,并实现自动分类。
[Abstract]:[objective] to study the automatic patent classification method for TRIZ application by constructing a personalized classification system. [methods] based on thematic model, a personalized classification system for TRIZ was constructed from macro, meso and micro levels. Through the combination of different classification features and algorithms, the combination with the highest classification accuracy is selected to construct the initial classifier, and the smooth non-equilibrium data and feature item reduction method are used to optimize the classifier. [results] A personalized classification system for TRIZ and a patent automatic classification based on this classification system are implemented. In the medium data level scenario (thousands), the patent automatic classification is realized, and the comprehensive evaluation index of classification effect is up to 90.2. [limitation] this method is not suitable for patent classification when the amount of data is small (100); in the case of large data volume (thousands), The validity of this method has not been verified. [conclusion] for the middle scale patent data, the classification system oriented to TRIZ can be constructed quickly and the automatic classification can be realized.
【作者单位】：中国科学院成都文献情报中心;中国科学院大学;
【基金】：中国科学院知识产权专项工作项目“中国科学院知识产权信息服务”(项目编号:KFJ-EW-STS-032) 中国科学院西部之光项目“基于本体的专利文献技术挖掘系统研究与实践”的研究成果之一
【分类号】：G306;G250.7

【相似文献】