Monday, October 25, 2010

Classification of Named Entities In Wikipedia

Fine Grained Classification of Named Entities In Wikipedia by Maksim Tkachenko, Alexander Ulanov, and Andrey Simanovsky has been published as HPL-2010-166.
This report describes the study on classifying Wikipedia articles into an extended set of named entity classes. We employed semi-automatic method to extend Wikipedia class annotation and created a training set for 15 named entity classes. We implemented two classifiers. A binary named-entity classifier decides between articles about named entities and other articles. A support vector machine (SVM) classifier trained on a variety of Wikipedia features determines the class of a named entity. Combination of the two classifiers helped us to boost classification quality and obtain classification quality that is better than state of the art.
Pretty technical, but anything that helps disambiguation sounds fine to me.

No comments: