A COMPARATIVE STUDY OF WORD SENSE DISAMBIGUATION OF ENGLISH MODAL VERB BY BP NEURAL NETWORK AND SUPPORT VECTOR MACHINE

作者：Yu Jianping^*; Huang Liang; Fu Jilin; Mei Deming

来源：International Journal of Innovative Computing Information and Control, 2011, 7(5A): 2345-2355.

• Word sense disambiguation
• Neural network
• Support vector machine
• Comparison

This paper applies back propagation neural network (BP NN) and support, vector machine (SVM) approaches in the word sense disambiguation (WSD) of English modal verb 'must' and compares the effects of WSD by the two models. First of all, a BP NN and a SVM for the WSD of English modal verb 'must' are established, respectively, and both reach an ideal correct disambiguation rate (98%). Then, based on the two models, a further investigation is carried out to see the influence of different features on the results of WSD of 'must'. After that, the two models are compared in. two aspects: (1) the performance in disambiguating root 'must' from epistemic `must'; (2) the performance in reflecting the influences of different linguistic (bag and relational) features on the effect of the WSD. The comparative results show that the SVM is more effective and has better generalization ability than, the BP NN; however, BP NN is more suitable for investigating the influence of individual linguistic feature on the effect of WSD than SVM. These comparative results provide very useful reference for model selection for WSD and for semantic studies.

• 出版日期2011-5
• 单位上海外国语大学; 燕山大学