HCKP: An Effective Hybrid Classification for Chinese News
Wanrong Gu · Journal of Information and Computational Science · 2014
Classifying massive Internet news is conducive to the subsequent process on the news applications. Traditional approaches strives to solve classification problem by learning the classification model based on word bag. However, the relationship between different classes, important factors in special corpus (e.g., topics, key persons, places or other factors in news article) are not addressed or solved efficiently in previous approaches. In order to solve these issues, we proposed a Hybrid Classification method based on the two-class classification method and the Key Persons (HCKP), in which we taken into account the relationship between different classes and set the key person factor into the model. Our experimental results show that by adjusting the weight of the key persons in the hybrid classification can get a good news classification performance compared to the state-of-the-art approaches in mass text data.