An Ensemble Algorithm for Sequential Labelling : A Case Study in Chemical Named Entity Recognition

Ensemble methods are learning algorithms that classify new data points by synthesizing the predictions of a set of classifiers. Many methods for constructing ensembles have been proposed such as weighted voting , manipulations of training samples, features, or labels. The paper proposes a novel e semble algorithm which constructs ensembles by manipulating the label set given to the learning algorithm and then classifies a new dataset by a voting algorithm specifically designed for sequential labelling task. The dataset released in th BioCreative V.5 CEMP (Chemical Entity Mention recognition) task was used to evaluate the performance of proposed algorithm. The results revea led th t the proposed algorithm can improve the precision and F-score.