Request Correction

Maximum Entropy Classifier Ensembling using Genetic Algorithm for NER in Bengali

In this paper, we propose classifier ensemble selection for Named Entity Recognition (NER) as a single objective optimization problem. Thereafter, we develop a method based on genetic algorithm (GA) to solve this problem. Our underlying assumption is that rather than searching for the best feature set for a particular classifier, ensembling of several classifiers which are trained using different feature representations could be a more fruitful approach. Maximum Entropy (ME) framework is used to generate a number of classifiers by considering the various combinations of the available features. In the proposed approach, classifiers are encoded in the chromosomes. A single measure of classification quality, namely F-measure is used as the objective function. Evaluation results on a resource constrained language like Bengali yield the recall, precision and F-measure values of 71.14%, 84.07% and 77.11%, respectively. Experiments also show that the classifier ensemble identified by the proposed GA based approach attains higher performance than all the individual classifiers and two different conventional baseline ensembles.