WebSets : Unsupervised Information Extraction approach to Extract Sets of Entities from the Web

We propose an unsupervised information extraction system, which exploits the structured information in the form of HTML tables to build meaningful sets of entities belonging to certain categories. Due to redundancy on the Web, we believe that entities belonging to important categories will frequently co-occur in table columns. We present a clustering algorithm to cluster such frequently occurring entities into meaningful sets. Then we find candidate category names for each set, using Hearst patterns like “X such as Y”, “X including Y”. Experimental results on four different datasets show that our method can extract meaningful sets of entities (with avg. cluster precision of 97-99%). It also proposes reasonable category names for them. We present an application of this method to enhance an existing knowledge base. Experiments show that our method improves the coverage of existing categories with 80-90% accuracy. It also suggests new categories that can be added to the knowledge base.