Using Wikipedia for Hierarchical Finer Categorization

Wikipedia is one of the largest growing structured resources on the Web and can be used as a training corpus in natural language processing applications. In this work, we present a method to categorize named entities under the hierarchical fine-grained categories provided by the Wikipedia taxonomy. Such a categorization can be further used to extract semantic relations among these named entities. More specifically, we examine instances of different kinds of Named Entities picked from Wikipedia articles categorized under 55 categories. We employ a Maximum Entropy based method to perform supervised learning that learns from local context of a named entity as well as a higher-level context such as hypernyms/hyponyms from Wikipedia and WordNet.