基于CRF和错误驱动的中心词识别

针对中文问题分类的中心词识别不准确的问题, 提出了一种基于条件随机场（CRF）和错误驱动学习相结合的识别方法。该方法采用CRF模型对问题的中心词进行初始标注, 依据词的上下文信息用错误驱动的学习方法对其标注结果进行纠正。在训练有序规则的过程中, 为了减少训练时间, 结合中心词的特点对错误驱动算法进行了改进。实验结果表明, 该方法在一定程度上提高了中心词的标注精度, 达到88%。

Recognition of focus based on CRF and TBL

Focusing on inaccurate recognition of focus in Chinese question classification, this paper presented a method based on condition random fields model (CRF) and TBL. First, it used the CRF model in tagging focuses in the questions initially. Then, it rectified the initial tagging results according to the contextual information with TBL algorithm. Besides, during the process of training ordered rules, it used an improvement on TBL algorithm based on the features of focus in order to reduce the training time. The experimental results show that the method can improve the accuracy rate of focus tagging to a great extend, and up to 88%.