Rbosaghz writesup Wang Cohen 2008

This is a review of Wang_2008_iterative_set_expansion_of_named_entities_using_the_web by user:Rbosaghz

This paper is about Set Expansion, which is directly tied to bootstrapping, a topic we've been looking at for the past few weeks. The task is to take a set of named entities and expand the set into more named entities of the same "type".

They present the SEAL system (Set Expansion for Any Language), which uses the web to expand lists. With their 2007 version, when given a large set of seeds (e.g., ten, which doesn't seem so many), SEAL's expansion method performs poorly. To fix this, the authors present Iterative SEAL (iSEAL), which allows the introduction of many seeds.

On every iteration the number of seeds is grown by a certain amount until a desired number of seeds is found. The authors consider and compare various rankings to pick the "best" candidate instances on each iteration. Their Random Walk (RW) ranking method seems to perform best in all experiments.