基于自学习规则和改进贝叶斯结合的问题分类*

根据对中文问题的分析可知，问题中的疑问词和中心词等关键词对问题所属类型起着决定性的作用。提出利用自学习方法建立疑问词—类别和疑问词+中心词—类别两种规则，并结合改进贝叶斯模型的问题分类方法。该方法充分利用了关键词对分类的贡献。实验结果表明，该分类方法有很大的改进，准确率达到了84%。

Question classification based on self-learning rules and modified Bayes

According to the Chinese question, this paper presented a question classification method which combined self-learning rules, consisting of question word-category rules and question word+head word-category rules established in advance by the self-learning method, and modified Bayesian model to improve Chinese question classification. At last, combined modified Bayesian model to improve question classification. The method takes advantage of the contribution of key words to Chinese question classification. Experimental results show that this classification method is a considerable improvement, and accuracy rate a chieves 84%.