Research on Column Concept Vector Based Web Table Matching

The Web consists of a huge number of structured data in the form of tables, which makes automatically integrating information from those tables of interest to ordinary users possible. A key problem of web table integration is the discovery of correspondences between web table columns. Most of traditional schema matching techniques can´t work well because of the lack of schema information and the small number of instance in the web tables. This paper presents a method of web table matching which is based on column concept vector. Column Heading Matcher and Instance Matcher are employed to enhance the matching accuracy. A set of experiments are applied to real-world web tables and the results demonstrate that our method has higher precision and accuracy.