Attribute Relation Extraction from Template-inconsistent Semi-structured Text by Leveraging Site-level Knowledge

A variety of methods have been proposed for attribute-value extraction from semistructured text with consistent templates (strict semi-text). However, when the templates in semi-structured text are inconsistent (weak semi-text), these methods will work poorly. To overcome the templateinconsistent problem, in this paper, we proposed a novel method to leverage sitelevel knowledge for attribute-value extraction. First, we use a graph-based random walk model to acquire site-level knowledge. Then we utilize such knowledge to identify weak semi-text in each page and extract attribute-value pairs. The experiments show that, comparing to the baseline method which does not utilize sitelevel knowledge, our method can improve the extraction performance significantly.