Identifying English BaseNPs Through a Combination of Maximum Entropy Approach and Brill Approach

To increase further the accuracy of BaseNP identification and utilize features of the maximum entropy approach and the Brill approach,an English BaseNPs identification algorithm based on a combined approach is presented.The algorithm is based on a high-performance POS(parts of speech) tagger.During the training phase and the application phase,maximum entropy approach is first applied to the initialization process of Brill approach,and the Brill approach is then run on its results already having high accuracy.Experimental results showed that this combined algorithm achieved a high precision and recall rate of over 94%,fully inosculating the strength of the maximum entropy approach and the Brill approach.It is comparable to the most ideal results of existing English BaseNP identification based on the same training and testing corpus.