An Ensemble Text Classification Model Combining Strong Rules and N-Gram

Most text classification methods apply machine learning methods, while ignored traditional rules-based methods. In this paper, we propose an ensemble text classification model combining classification rules and N-gram language model. In order to generate strong classification rules, we propose an exhaustive noun-phrase extraction algorithm and a new optimized rule induction algorithm, called SCA (strong covering algorithms). We also introduce an improved good-Turing (GT) smoothing method for N-gram model. Experimental results show that our ensemble classifier achieves an approximately 8% improvement as compared to bi-gram with word-based classifier and 15% improvement as compared to traditional rules-based classifier. In conclusion, the accuracy of ensemble classification model is competent to only a single classification models.