Entity answer extraction of web table

This paper presents an entity answer extraction method based on list web table. Firstly, extract table from page using the features of web page table and label, segment the table that includes the potential entity answers by calculating the relevance of web table´s title and query context, merge the table elements of each column according to table properties, and merge the web table´s title with the merged elements of column again. Secondly, using merged passage as context of entity recognition, and recognize the entity for each element of the table, thus get the probability of the column elements belongs to the same type of entity answer, and locate the passages of entity answers and the entity answers. Finally, we conduct the experiment in the task of Entity Track of TREC2009. It turns out that the proposed method shows a very good result, the accuracy of entity answer extraction for web table has achieved 99.08%.