自动答疑系统中问题定位方法的研究
摘要
自动问答系统自二十世纪中期出现以来一直处于快速发展之中,成为计算机领域内的研究热点。问题定位是问答系统中关键技术之一,目前大多数问答系统采用为答案库建立索引的方式来实现问题定位,达到缩小查找范围的目的。然而这些方法都存在一定的问题。本文针对远程教育特定领域的自动答疑系统,提出了利用模糊自适应谐振映射神经网络(Fuzzy ARTMAP)以及支持向量机(SVM)的方法解决问题定位。重点研究和分析问题的特征表示以及问题的定位方法。
论文首先介绍了自动问答系统的发展过程、一般体系结构、常用技术和定位方法的研究现状;重点研究了答案库的结构设计和预处理、问题的特征以及面向高维空间的特征变换,并结合Fuzzy ARTMAP和SVM两个分类算法分析了基于特征表示与变换的问题定位方法;最后,给出了原型的设计与实现,并对实验的性能进行了分析,验证了论文技术的可行性。
QA system has been developing rapidly from the mid of 20th century, and has become the hotspot of research in computer science area. Localization is one of key technique in QA system. Most QA systems create index for answers’collection to reduce searching range and localize, however these methods have certain weaknesses. Aiming at QA system in remote education background, this paper proposes a new method using Fuzzy ARTMAP and SVM to solve localization. This paper’s background is based on a certain subject QA system, and focuses on question feature representation and question’s localization.
This paper firstly introduces the development, architecture and techniques of QA system and the status of localization, secondly focuses on the pretreatment of answers’collection, questions’feature representation and feature transformation to high dimension space, thirdly, analyses the localization method of questions with SVM and Fuzzy ARTMAP; at last, designs and implements an prototype and analyses the performance to verify the feasibility of arithmetic proposed by this paper .
论文首先介绍了自动问答系统的发展过程、一般体系结构、常用技术和定位方法的研究现状;重点研究了答案库的结构设计和预处理、问题的特征以及面向高维空间的特征变换,并结合Fuzzy ARTMAP和SVM两个分类算法分析了基于特征表示与变换的问题定位方法;最后,给出了原型的设计与实现,并对实验的性能进行了分析,验证了论文技术的可行性。
QA system has been developing rapidly from the mid of 20th century, and has become the hotspot of research in computer science area. Localization is one of key technique in QA system. Most QA systems create index for answers’collection to reduce searching range and localize, however these methods have certain weaknesses. Aiming at QA system in remote education background, this paper proposes a new method using Fuzzy ARTMAP and SVM to solve localization. This paper’s background is based on a certain subject QA system, and focuses on question feature representation and question’s localization.
This paper firstly introduces the development, architecture and techniques of QA system and the status of localization, secondly focuses on the pretreatment of answers’collection, questions’feature representation and feature transformation to high dimension space, thirdly, analyses the localization method of questions with SVM and Fuzzy ARTMAP; at last, designs and implements an prototype and analyses the performance to verify the feasibility of arithmetic proposed by this paper .
引文
[1]Ricardo Baeza-Yates, Berthier Ribeiro-Neto 等著。现代信息检索(英文版)[M]。北京:机械工业出版社,2004.2
[2]邓乃扬,田英杰. 数据挖掘中的新方法——支持向量机[M]. 北京:科学出版社,2004.6
[3]Nello Cristianini, John Shawe-Taylor 著,支持向量机导论[M]. 李国正,王猛,曾华军译. 北京:电子工业出版社,2004.3
[4]John Shawe-Taylor,Nello Cristianini.模式分析的核方法(英文版)[M]. 北京:机械工业出版社,2005.1
[5]剑锋,卜东波,白硕.基于向量空间模型的文本自动分类系统的研究与实现[J/OL]. http://www.keenage.com/html/c_index.html
[6]Carpenter,G.A.,Grossberg,S.The ART of adaptive pattern recognition by a self-organizing neural network[J]. Digital Object Identifier,1988,21(3):77 - 88
[7]Carpenter, G.A., Grossberg, S., & Rosen, D.B.ART 2-A: An adaptive resonance algorithm for rapid category learning and recognition[J]. Neural Networks,1991,4: 493-504.
[8] Gail A. Carpenter, Stephen Grossberg, and David B. Rosen. Fuzzy ART: Fast stable learning and categorization of analog patterns by an adaptive resonance system[J]. Neural Networks, 4:759-771, 1991.
[9]Carpenter, G.A., Gopal, S., Shock, B.M., & Woodcock, C.E.ARTMAP neural network classification of land use change[J]. Proceedings of the World Congress on Computers in Agriculture and Natural Resources, Igua?a Falls, Brazil, September, 2001.
[10]刘群,李素建.基于《知网》的词汇语义相似度计算[J/OL]. http://www.keenage.com/html/c_index.html
[11]蕫振东,蕫强.“知网”[J/OL].http://www.keenage.com
[12] Bart Kosko著,模糊工程[M]. 黄崇福译.西安:西安交通大学出版社,1999.6
[13] Kuh, A. Adaptive least square kernel algorithms and applications[J].Neural Networks, 2002. IJCNN’02 Proceedings of the 2002 International Joint 2002,3(12):2104 - 2107
[14]Takahashi, N.; Nishi, T.Rigorous proof of termination of SMO algorithm for support vector Machines[J].Neural Networks, 2005,16(3) :774 – 776
[15]Christos Christodoulou, Michael Georgiopoulos.Applications of Neural Networks in Electromagnetics[M], Artech House Publishers,2001.1
[16]张林,胡波.基于主元分析和FuzzyAR7,模型的人脸识别算法[J].电路与系统学报,1999, 14 (3): 9-16
[17]韩敏,程磊,唐晓亮。Fuzzy ARTMAP 神经网络在土地覆盖分类中的应用研究[J]。中国图像图形学报,2005,10(4):415-419
[18]王能斌. 数据库系统原理[M]. 北京:电子工业出版社,2000.4
[19]徐易. 智能答疑系统的研究与实现:[学位论文][D] . 东南大学计算机系,2003
[20]Yi Guan,Xiao-Long Wang.Quantifying semantic similarity of Chinese words from HowNet. Machine Learning and Cybernetics,2002. Proceedings. 2002 International Conference,2002,1:234 -239
[21]Carpenter,G.A., Grossberg,S., & Reynolds, J.H.ARTMAP: Supervised real-time learning and classification of nonstationary data by a self-organizing neural network[J]. Neural Networks, 4, 565-588,1991.
[22]Carpenter,G.A. & Grossberg, S.ART 3: Hierarchical search using chemical transmitters in self- organizing pattern recognition architectures[J].Neural Networks, 3, 129-152, 1990
[23]Carpenter, G.A.Default ARTMAP[J].Proceedings of the International Joint Conference on Neural Networks (IJCNN'03), Portland, Oregon,2003.
[24]程云鹏等.矩阵论(第二版).西安:西北工业大学出版社[M],1999 Martin T.Hagan,Howard B.Demuth,Mark Beale.神经网络设计(英文影印版)[M]. 北京:机械工业出版社,2002.9
[25]朱衡君,肖燕彩,邱成.MATLAB 语言及实践教程[M].清华大学出版社,2005.1
[26]飞思科技产品研发中心.MATLAB 6.5 辅助神经网络分析与设计[M].北京:电子工业出版社 ,2003.1
[27]Carpenter GA, Grossberg S, Markuzon N, et al. FuzzyARTMAP: Aneural network architecture for incremental supervised learning ofanalog multidimensional map s [J]. NeuralNetworks, 1992, 3(5) : 698～713.
[28]丁月华,文桂华,郭炜强。基于核向量空间模型的专利分类[J]。华南理工大学学报,2005,33(8):58-61
[29]Song-Feng Zheng. Least Square Support Vector Machine and Its Bayesian Interpretation[EB/OL]. http://www.stat.ucla.edu/~sfzheng/Courses/LSSVM_Bayesian.doc
[30]Thorsten Joachims .Text Categorization with Support Vector Machines: Learning with Many Relevant Features. Proceedings of ECML-98, 10th European Conference on Machine Learning, 1997
[31]Thorsten Joachims.Learning to Classify Text Using Support Vector Machines : Methods, Theory and Algorithms (The International Series in Engineering and Computer Science) [M].Springe,2002.4
[32]C. Cortes and V. Vapnik. Support-vector networks. Machine Learning,1995,20:273-297.
[2]邓乃扬,田英杰. 数据挖掘中的新方法——支持向量机[M]. 北京:科学出版社,2004.6
[3]Nello Cristianini, John Shawe-Taylor 著,支持向量机导论[M]. 李国正,王猛,曾华军译. 北京:电子工业出版社,2004.3
[4]John Shawe-Taylor,Nello Cristianini.模式分析的核方法(英文版)[M]. 北京:机械工业出版社,2005.1
[5]剑锋,卜东波,白硕.基于向量空间模型的文本自动分类系统的研究与实现[J/OL]. http://www.keenage.com/html/c_index.html
[6]Carpenter,G.A.,Grossberg,S.The ART of adaptive pattern recognition by a self-organizing neural network[J]. Digital Object Identifier,1988,21(3):77 - 88
[7]Carpenter, G.A., Grossberg, S., & Rosen, D.B.ART 2-A: An adaptive resonance algorithm for rapid category learning and recognition[J]. Neural Networks,1991,4: 493-504.
[8] Gail A. Carpenter, Stephen Grossberg, and David B. Rosen. Fuzzy ART: Fast stable learning and categorization of analog patterns by an adaptive resonance system[J]. Neural Networks, 4:759-771, 1991.
[9]Carpenter, G.A., Gopal, S., Shock, B.M., & Woodcock, C.E.ARTMAP neural network classification of land use change[J]. Proceedings of the World Congress on Computers in Agriculture and Natural Resources, Igua?a Falls, Brazil, September, 2001.
[10]刘群,李素建.基于《知网》的词汇语义相似度计算[J/OL]. http://www.keenage.com/html/c_index.html
[11]蕫振东,蕫强.“知网”[J/OL].http://www.keenage.com
[12] Bart Kosko著,模糊工程[M]. 黄崇福译.西安:西安交通大学出版社,1999.6
[13] Kuh, A. Adaptive least square kernel algorithms and applications[J].Neural Networks, 2002. IJCNN’02 Proceedings of the 2002 International Joint 2002,3(12):2104 - 2107
[14]Takahashi, N.; Nishi, T.Rigorous proof of termination of SMO algorithm for support vector Machines[J].Neural Networks, 2005,16(3) :774 – 776
[15]Christos Christodoulou, Michael Georgiopoulos.Applications of Neural Networks in Electromagnetics[M], Artech House Publishers,2001.1
[16]张林,胡波.基于主元分析和FuzzyAR7,模型的人脸识别算法[J].电路与系统学报,1999, 14 (3): 9-16
[17]韩敏,程磊,唐晓亮。Fuzzy ARTMAP 神经网络在土地覆盖分类中的应用研究[J]。中国图像图形学报,2005,10(4):415-419
[18]王能斌. 数据库系统原理[M]. 北京:电子工业出版社,2000.4
[19]徐易. 智能答疑系统的研究与实现:[学位论文][D] . 东南大学计算机系,2003
[20]Yi Guan,Xiao-Long Wang.Quantifying semantic similarity of Chinese words from HowNet. Machine Learning and Cybernetics,2002. Proceedings. 2002 International Conference,2002,1:234 -239
[21]Carpenter,G.A., Grossberg,S., & Reynolds, J.H.ARTMAP: Supervised real-time learning and classification of nonstationary data by a self-organizing neural network[J]. Neural Networks, 4, 565-588,1991.
[22]Carpenter,G.A. & Grossberg, S.ART 3: Hierarchical search using chemical transmitters in self- organizing pattern recognition architectures[J].Neural Networks, 3, 129-152, 1990
[23]Carpenter, G.A.Default ARTMAP[J].Proceedings of the International Joint Conference on Neural Networks (IJCNN'03), Portland, Oregon,2003.
[24]程云鹏等.矩阵论(第二版).西安:西北工业大学出版社[M],1999 Martin T.Hagan,Howard B.Demuth,Mark Beale.神经网络设计(英文影印版)[M]. 北京:机械工业出版社,2002.9
[25]朱衡君,肖燕彩,邱成.MATLAB 语言及实践教程[M].清华大学出版社,2005.1
[26]飞思科技产品研发中心.MATLAB 6.5 辅助神经网络分析与设计[M].北京:电子工业出版社 ,2003.1
[27]Carpenter GA, Grossberg S, Markuzon N, et al. FuzzyARTMAP: Aneural network architecture for incremental supervised learning ofanalog multidimensional map s [J]. NeuralNetworks, 1992, 3(5) : 698～713.
[28]丁月华,文桂华,郭炜强。基于核向量空间模型的专利分类[J]。华南理工大学学报,2005,33(8):58-61
[29]Song-Feng Zheng. Least Square Support Vector Machine and Its Bayesian Interpretation[EB/OL]. http://www.stat.ucla.edu/~sfzheng/Courses/LSSVM_Bayesian.doc
[30]Thorsten Joachims .Text Categorization with Support Vector Machines: Learning with Many Relevant Features. Proceedings of ECML-98, 10th European Conference on Machine Learning, 1997
[31]Thorsten Joachims.Learning to Classify Text Using Support Vector Machines : Methods, Theory and Algorithms (The International Series in Engineering and Computer Science) [M].Springe,2002.4
[32]C. Cortes and V. Vapnik. Support-vector networks. Machine Learning,1995,20:273-297.