Volume 1 (1) 2007
|Title:||A Methodology for Feature Selection in Named Entity Recognition|
|Authors: ||Fredrick Edward Kitoogo and Venansius Baryamureeba|
|Published:|| ©IJCIR Vol1 (1) 2007, PP. 18-26|
In this paper a method for feature selection in named entity recognition is proposed. Unlike traditional named entity recognition approaches which mainly consider accuracy improvement as the sole objective, the innovation here is manifested in the use of a multi-objective genetic algorithm which is employed for feature selection basing on various aspects including error rate reduction and time taken for evaluation, and also demonstrating the use of Pareto optimization. The proposed method is evaluated in the context of named entity recognition, using three different data sets and a K-nearest Neighbour machine learning algorithm. Comprehensive experiments demonstrate the feasibility of the methodology.
View full Article
|General Terms: || Computer Science, Language Processing Additional Key Words and Phrases: feature selection, multi-objective genetic algorithm, named entity recognition.|
|Categories and Subject Descriptors: || I.5.2 [Pattern Recognition]: Design MethodologyóFeature evaluation and selection; I.2.7 [Artificial Intelligence]: Natural Language processingólanguage parsing and understanding|