基于字角色标注的中文书目关键词标引研究

Research on Keywords Indexing for Chinese Bibliography Based on Word Roles Annotation

中文书目机器自动标引是数字图书馆建设中亟待解决的关键问题之一。本文试图将条件随机场(CRFs)序列标注机器学习算法引入到关键词抽取中,建立面向图书内容、基于字角色标注的中文书目关键词标引模型。将图书内容转化为字序列,进而提出构建关键词角色空间模型和综合利用字序列上下文特征的设计思路。通过实验,从题名和内容提要中分别自动抽取关键词,论证该模型的合理性和实用性。

Automatic indexing by computers for Chinese bibliography has become one of the most critical problems which should be solved immediately in digital library construction. This paper tries to introduce Conditional Random Fields （CFRs） algorithm into the keyword extraction of Chinese bibliography, and builds the model which faces book contents based on the word roles annotation. The model turns the book contents into sequences of words. Based on that, an idea which combines word roles space model building with context features of word sequence comprehensive u- tilization has been proposed. Moreover, the paper also verifies the rationality and practicality of the model by showing the experiment of automatically extracting keywords from titles and abstracts. 6 figs. 3 tabs. 23 refs.