基于IPC知识结构的专利自动分类模型  

Automated Categorization Model of Patent Based on the Knowledge of IPC

作　　者：刘玉琴[1] 赖院根[1] 雷孝平[1] 

机构地区：[1]北京理工大学管理与经济学院,北京100081

出　　处：《小型微型计算机系统》2007年第12期2295-2298,共4页Journal of Chinese Computer Systems

基　　金：国家自然科学基金重点项目(70031010)资助;"九八五工程"国防科技管理与国防动员研究中心资助

摘　　要：构建的专利自动分类模型,利用国际专利分类号自身的类别信息建立类别特征词原始集合,结合现有专利进行扩充训练.计算待分类专利中所有类别的特征词频率向量,进而判断专利与各类别的关联程度,实现专利的自动分类.实验结果显示,该模型的分类效果在大类、小类层次上较好.Using hierarchical structure information of IPC, automated categorization model of patent developed in the paper constructed original sets of class character words, and then extension training is done drawing on the patents that have been given correct IPC codes. The model calculated all classes＇ frequency vectors of character words in patent and the correlation degree between patent and IPC classes, so decided the class of the patent. Experimental results show that the model works well on the level of class and subclass.

关 键 词：文本分类 专利分类 国际专利分类号 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]