Towards a high quality and web-scalable table search engine

Abstract

For over a decade, a large number of studies have explored efficient mechanisms for finding relevant information from structured sources such as relational, semi-structured, and graph databases. While successful in its own right, keyword search over structured data has yet to gain wide spread adoption on the Web, and it is not because of the lack of structured data on the Web. In fact, the Web offers orders of magnitude more structured data than any offline data source: 14 billions tables can be gathered just by considering page content between the table tags. The main reason is that keyword search over structured data on the Web presents a unique set of challenges that are quite different from its non-Web counterparts, as well as different from searching over documents (where the search engines have excelled). In this talk, I will discuss those challenges and our approaches in addressing them at Google's WebTables project. In particular, I will present the table search engine, our initial effort toward building a high quality and scalable structured data search engine, along with our other efforts on managing structured data on the Web.