Abstract: Web content extraction has great engineering and application value in the fields of information retrieval, text analysis and network resource data processing. In view of the problem of web content extraction caused by useless information on web pages and the heterogeneity of web page structures, this paper proposes an automated web page content extraction method based on Document Object Model (DOM). Firstly, for DOMs generated from original web pages, we remove useless nodes from them and then compress the models, which facilitates subsequent processing. Then, we identify the web page content based on text and hyperlink density. Finally, we identify the noise hyperlinks based on node entropy and remove them from the content. The experimental results show that compared with the traditional methods of web page content extraction, the accuracy and F1 score of our method are obviously improved while there is only a slight decline on recall.

Key words: document object model (DOM), content extraction of web pages, text density, node entropy