Bbd writeup of Brin 1999

This is a review of Brin_1999_extracting_patterns_and_relations_from_the_world_wide_web by user:Bbd.

This paper presents a technique which leverages duality between sets of patterns and relations to bootstrap the set of target relations starting from a small sample. He describes a fairly simple technique to do it in iterative way.

Step 1 : From available relation instances find out the patterns matching them. Step 2 : From the patterns found in Step 1, extract more relation instances.

This bootstraps the system nicely. He presents an example of bootstrapping (author,book title) relation and starting from just 5 sample instances, he could found 15K instances of author,book titles. I liked this paper because it is very simple, intuitive technique for bootstrapping the relation instance set.