Layered and Weighted Tree Matching Algorithm for Automatic Web Data Records Recognition

Automated web data record analysis and recognition is an important issue for improving the automation of web information extraction. Some typical methods perform similar data record analysis based on Simple Tree Matching (STM) algorithm. However, STM assigns elements the same weight and neglects the different impacts of different types of elements. Thus much noise would be induced during data record analysis, which eventually decreases the precision of data record recognition. This paper proposes a Layered and Weighted Tree Matching (LWTM) algorithm. First, we propose a layered filtering strategy to filter out potential noise elements. Then, we propose a weighted tree matching algorithm which assigns different weights for different types of HTML elements in terms of their importance for data record analysis. Further we combine the layered filtering strategy and the weighted tree matching algorithm to further improve the analysis results. Experimental results show that the proposed LWTM algorithm outperforms the STM-based methods.