Attribute Reduction for Chinese Question Classification

作者：Yuan Liwei; Su Lei^*; Shu Peng

来源：28th Chinese Control and Decision Conference, 2016-05-28 To 2016-05-30.

DOI：10.1109/CCDC.2016.7531977

• Community-based Question Answering
• Question classification
• rough set
• Extract Features
• Attribute reduction

With the rapid development of the question and answer services based on community, like Sina Ask, Baidu Zhidao and Yahoo! The Community-based Question Answering service has been became a new knowledge-sharing model with characteristics of interactivity and openness. The community sites provide high quality service to meet clients' need and attract them actively participation. In order to accurately understand the user query and provide useful information, it is necessary to deal with the questions in the community. Additionally, question classification in the CQA is the key component in this step. As everyone knows, the difficulty of the question classification is High-dimensional feature vector, usually uses feature selection as the primary method of dimensionality reduction, The author adopts the following way in this paper: Firstly, we usually use feature selection to screen out first feature subset, then use the dependence of the rough set as a heuristic information to deeply feature selection, last, screen out more useful features in the experiment. The experiment data suggests that the algorithm is effective, it can not only reduce the dimension of questions, but also improve the classification accuracy effectively.

• 出版日期2016
• 单位昆明理工大学