摘要
在解析微博文本语法的基础上,结合实体关系的定义和形式化表示,提出了采用关系网络有向图模型的方法来反映文本之间的结构关系,较好地表达了文本的语义信息,弥补了词频特征刻画的不足之处。利用改进后的TPR(Topic-PAGERANK)测算各节点对应的度来表现关系元组的重要程度,按序输出关系元组对应的原博文语义字段作为摘要。最后,通过实验证明了基于关系网络的文本自动文摘方法抽取出的摘要涵盖信息更全面,冗余更少。
On the basis of syntax parsing, combining the definition of entity relationship and formalized representation, this paper put forward a method based on directed graph model to reflect the structured relationship between texts, ex- pressing text semantic information, making up for the shortcomings of word frequency characteristics. After that, the corresponding value of each node is measured with improved TPR （Topic-PAGERANK） to represent the importance of the relationship group. Then the corresponding original microblog text of relational tuples is sequentially outputed. Fi- nally, it is proved by experiments that the text summarization extracted by automatic text summarization method based on relational tuDle is more comprehensive and less redundant.
出处
《计算机科学》
CSCD
北大核心
2016年第9期77-81,共5页
Computer Science
基金
国家863高技术研究发展计划(2012AA01A401)资助
