Column concept determination based on multiple evidences

Tables on the web provide rich information. To make sufficient usage of web tables, the semantics of columns should be identified correctly. The absence, misspelling, and abbreviation in column names bring the challenges in column semantics identification. Facing this challenge, we extract multiple features including keywords, concepts, and structure from the content in the column. Thus, we could identify the column semantics by matching these multiple features. For the extraction and matching with these features, we propose efficient algorithms. Experimental results on real data sets show that our solution achieves high performance.