一种基于粗糙集增量式规则学习的问题分类方法研究

李鹏; 王晓龙; 关毅

doi:10.3724/SP.J.1146.2006.01689

一种基于粗糙集增量式规则学习的问题分类方法研究

doi: 10.3724/SP.J.1146.2006.01689

基金项目:

国家自然科学基金重点项目(60435020)、国家自然科学基金项目(60504021)和国家863目标导向类课题(2006AA01Z197)资助课题

计量
- 文章访问数: 3384
- HTML全文浏览量: 89
- PDF下载量: 1212
- 被引次数: 0
出版历程
- 收稿日期: 2006-10-30
- 修回日期: 2007-05-21
- 刊出日期: 2008-05-19

Question Classification with Incremental Rule Learning Algorithm Based on Rough Set

摘要

摘要: 该文提出一种基于粗糙集增量式规则自动学习来实现问题分类的方法,通过深入提取问句特征并采用决策表形式构建训练语料,利用机器学习的方法自动获取分类规则。与其他方法相比优势在于,用于分类的规则自动生成,并采用粗糙集理论的简约方法获得优化的最小规则集;首次在问题分类中引入增量式学习理念,不但提高了分类精度,而且避免了繁琐的重新训练过程,大大提高了学习速度,并且提高了分类的可扩展性和适应性。对比实验表明,该方法分类精度高,适应性好。在国际TREC2005Q/A实际评测中表现良好。
- 粗糙集; 问题分类; 增量式学习; 决策表; 特征选择
Abstract: This paper presents a method on automatic question classification through incremental rule learning based on rough set theory. The core of the method is appling the machine learning approach to gain classified rules automatically through extract the features of query sentence thoroughly, and the decision table is used to construct the training collection. Comparing with the alternative means, the superiority is that it acquires the classified rule automatically and uses the rough set method to obtain the optimized smallest rule set. Especially, the incremental learning is induced to improve the precision and avoid the tedious re-training process. The performance of the approach is promising, when tested on opposite test. Meanwhile, the method obtains a very good result in the international TREC2005 Q/A track.

HTML全文

参考文献(1)

Marius A Pasca. High-performance, open-domain questionanswering from large text collections. [Ph. D. dissertation],University of Southern Methodist, 2001.[2]Cody Kwok, Oren Etzioni, and Daniel. Scaling questionanswering to the web [J]. ACM Trans. on InformationSystems, 2001, 9(3): 242-262.[3]Shaw M L G and Gaines B R. Question classification inrule-based systems [C]. Proceedings of Expert Systems86,The 6Th Annual Technical Conference on Research anddevelopment in expert systems, Brighton, 1987: 123-131.[4]张宇, 刘挺. 基于改进贝叶斯模型的问题分类 [J]. 中文信息学报, 2005, 19(2): 100-105.Zhang Yu, Liu Ting, and Wen Xu. Modified Bayesian modelbased question classification [J]. Journal of ChineseInformation Processing, 2005, 19(2): 100-105.[5]Taira Jun Suzuki, Sasaki Yutaka, and Maeda Eisaku.Question classification using HDAG kernel [C]. ACLWorkshop on Mulitilingual Summarization and QuestionAnswering, Sapporo, 2003: 61-68.[6]Li Xin and Roth Dan. Learning question classifier [C]. InProceedings of the 19th International Conference onComputational Linguistics (COLING02). Taipei, 2002:556-562.[7]Zhang Dell and Lee Wee Sun. Question classification usingsupport vector machines [C]. Proceedings of the 26th annualinternational ACM SIGIR Conference on Research andDevelopment in Information Retrieval, New York, ACMPress, 2003: 26-32.[8]王国胤. Rough 集理论与知识获取[M]. 西安：西安交通大学出版社.[J].2001.Wang Guo-yin. Rough Sets Theory and KnowledgeDiscovery[M]. Xian Jiaotong University Press.2001,:-[9]王国胤, 于洪等. 基于条件信息熵的决策表约简 [J]. 计算机学报, 2002, 25(7): 759-766.Wang Guo-yin and Yu Hong. Decision table reduction basedon conditional information entropy [J]. Chinese J Computer,2002, 25(7): 759-766.[10]于洪，杨大春，吴中福. 基于 Rough set 理论的增量式规则获取算法[J]. 小型微型计算机系统, 2005, 26(1): 36-41.Yu Hong, Yang Da-chun, and Wu Zhong-fu. Incremental ruleacquisition algorithm based on rough set [J]. Mini-MicroSystems, 2005, 26(1): 36-41.[11]Pawlak Z. Rough set: theoretical aspects and reasoning aboutdata [M]. Dordrecht, Kluwer Academic Publishers, 1991.[12]Ellen M. Voorhees, and Hoa Trang Dang. Overview of theTREC 2005 question answering Track [C]. The FourteenthText REtrieval Conference (TREC 2005) Proceedings. NewYork, 2005: 1-15.

施引文献

资源附件(0)

访问统计