点击次数：

论文类型： 期刊论文

发表时间： 2009-06-01

发表刊物： Journal of Information and Computational Science

收录刊物： EI、Scopus

卷号： 6

期号： 3

页面范围： 1495-1503

ISSN号： 15487741

摘要： We firstly introduce the model of community identification to keywords extraction in web pages. Max- Flow algorithm can be used to identify a community in a local web graph which is concentrated on one topic. A web page contains a single relatively extensive topic too. Based on this observation, we treat words in a web page as nodes and relations between words as edges to construct a graph, rank words in the graph to find out some very important words as Seed-Keywords, and then input the graph and the Seed-Keywords to a modified version of Max-Flow algorithm to output a community, whose members are viewed as Target-Keywords. In this process, we do word sense disambiguation in a kind of context 'Topic-Block', whose precision is compared with a coarser-grained context, the whole web page and a finer-grained context, the basic element in HTML. The experiment results show that Topic-Block based word sense disambiguation is effective and Max-Flow algorithm can extract any number keywords adaptive to the size of web pages. 1548-7741/ Copyright ? 2009 Binary Information Press.