基于语用情境的资源推荐研究及应用
|
摘要
随着互联网的飞速发展,网络信息过载已成为目前网络用户所面临的主要问题,资源推荐系统为解决这一问题提供了有效手段,它可以为用户提供信息过滤和资源推荐服务,提高了用户的工作效率,正逐渐被大多数用户所青睐,而资源推荐的研究亦成为学者研究的重要领域。目前,现有的主要推荐系统有基于规则的推荐、基于内容的推荐、协同过滤推荐等。
基于规则的推荐是根据用户和规则模型产生的推荐,它可以满足用户实时性的需要,但规则的制定需要领域专家的参与,随着时间的推移会产生偏差和难以更新的问题,从而降低推荐的效率。基于内容的推荐是通过计算被推荐资源的内容与用户兴趣的相似性来选择所要推荐的资源,但是它只能依靠用户的兴趣进行推荐,并不能找到用户新的兴趣之处,所以推荐也会局限在用户访问的历史记录中。
协同过滤推荐目前是资源推荐系统中最成功的资源推荐技术,并且在很多的资源推荐系统中都得到了大量的应用,协同过滤推荐是从用户出发,寻找目标用户的最近邻居用户,利用最近邻居对其他资源的加权评价值作为目标用户对该资源的评价,并以此为依据,向目标用户进行资源推荐,所以它可以为用户发现潜在的兴趣。虽然协同过滤推荐得到了成功的应用,并且具有很多优势,但传统的协同推荐都是通过用户对项目的评分作为基础,计算目标用户的相似邻居,然后为用户推荐资源,这样就造成了数据评分矩阵稀疏以及冷启动的问题,并且评分也无法完全反映出用户的兴趣爱好。
针对以上问题,本文提出了基于语用情境的推荐方法。在传统的协同推荐方法的基础上,将语用学中的情境因素和信任因素引入到资源推荐的研究中来,其中,影响用户行为的上下文信息和场景信息成为用户情境,对用户决策产生作用的用户情境成为最显著情境,该方法首先获取用户的最显著个性化情境因素,然后结合用户最显著情境计算出用户兴趣相关性、评分相似性和评分相关性矩阵,形成用户兴趣关联模型,最后计算出用户信任等级,以用户信任等级值作为用户的评分权重,结合传统的推荐方法预测目标用户的评分。
本文采用Matlab在MovieLens数据集上对该方法进行了仿真实验,实验数据表明,该方法具有可行性和有效性,较之传统的协同推荐方法和Slope One算法在平均绝对误差上有一定的提高,为现有的资源推荐技术提供了一种新的方向。最后,本文结合作者参与的高教社基于本体、语义和语用的智能化教育资源平台项目,设计了一个基于用户情境的资源推荐系统模型,以此来说明该方法在实际中的应用。
With the rapid advancement of the Internet, the information overload has been a primal problem of Internet users. Resource recommendation system provides a very effective means of solving this problem. It can provide users with information filtering and the service of resource recommendation to improve their work efficiency, which is gradually welcomed by most users. So the research of resource recommendation is becoming a vital area. Presently, the main resource recommendation systems are system based on rules, one on contents and collaborative filtering recommendation.
System based on rules is formed from user and rule model. It can meet users’real-time needs, but the formulation of rules requires the involvement of experts, and the efficiency will be lowered, for the emergence of deviation and problems that is difficult to update. System based on contents chooses what is going to be recommended according to the contents of the information and the user interest. The shortcoming is that it relays only on user interest, and cannot find where the new interest of user. Therefore, its recommendation is restricted in the area of history list.
Among current resource recommendation systems, the collaborative filtering recommendation is the most successful one and extensively used in a number of resource recommendation systems. Collaborative filtering recommendation starts from user to seek the nearest neighbor user of the object user. By evaluating the other resource’s weighted evaluation value of the nearest neighbor as the target user’s expectation, it provided the resources to user, So the potential interest of user will be found. Though with a good many advantages, collaborative filtering recommendation has been successfully applied, the traditional collaborative filtering recommendation based on the project grade calculates similar neighbors of the object user then recommends resource. This will cause data sparsity and cold start, and the interest of users can not be fully reflected by the grade.
To settle problem above, the paper proposes a recommendation system based on pragmatics. On the basis of the traditional collaborative filtering recommendation, introduce the situation factor and factors of trust of pragmatics into the research of resource recommendation. Among all these information, the contextual information and scene information that influence user’s behaviors turn to user scenarios, and the user scenarios that militates user decision turns to the most prominent scenarios. This method acquires user’s prominent scenarios in the beginning, combines them to calculates user interest relativity, rating similarity and rating correlation matrix, then forms the correlation model of user interest, finally, calculates user confidence grade. Taking the user confidence grade as user score weight, combine the traditional method to predict the grade of object user.
This paper adopts Matlab and dose an simulation experiment in MovieLens. The experiment data shows that the method has a certain feasibility and effectiveness, and compared with the traditional one, the Slope One algorithm can reduce absolute average error. And it provides a new direction for current resource recommendation technology. Finally, combining with the intelligent educational platform project of Higher Education Press based on ontology, semantic and pragmatic, the paper designs a resource recommendation model based on user learning situation to explain the practical application of this method.
基于规则的推荐是根据用户和规则模型产生的推荐,它可以满足用户实时性的需要,但规则的制定需要领域专家的参与,随着时间的推移会产生偏差和难以更新的问题,从而降低推荐的效率。基于内容的推荐是通过计算被推荐资源的内容与用户兴趣的相似性来选择所要推荐的资源,但是它只能依靠用户的兴趣进行推荐,并不能找到用户新的兴趣之处,所以推荐也会局限在用户访问的历史记录中。
协同过滤推荐目前是资源推荐系统中最成功的资源推荐技术,并且在很多的资源推荐系统中都得到了大量的应用,协同过滤推荐是从用户出发,寻找目标用户的最近邻居用户,利用最近邻居对其他资源的加权评价值作为目标用户对该资源的评价,并以此为依据,向目标用户进行资源推荐,所以它可以为用户发现潜在的兴趣。虽然协同过滤推荐得到了成功的应用,并且具有很多优势,但传统的协同推荐都是通过用户对项目的评分作为基础,计算目标用户的相似邻居,然后为用户推荐资源,这样就造成了数据评分矩阵稀疏以及冷启动的问题,并且评分也无法完全反映出用户的兴趣爱好。
针对以上问题,本文提出了基于语用情境的推荐方法。在传统的协同推荐方法的基础上,将语用学中的情境因素和信任因素引入到资源推荐的研究中来,其中,影响用户行为的上下文信息和场景信息成为用户情境,对用户决策产生作用的用户情境成为最显著情境,该方法首先获取用户的最显著个性化情境因素,然后结合用户最显著情境计算出用户兴趣相关性、评分相似性和评分相关性矩阵,形成用户兴趣关联模型,最后计算出用户信任等级,以用户信任等级值作为用户的评分权重,结合传统的推荐方法预测目标用户的评分。
本文采用Matlab在MovieLens数据集上对该方法进行了仿真实验,实验数据表明,该方法具有可行性和有效性,较之传统的协同推荐方法和Slope One算法在平均绝对误差上有一定的提高,为现有的资源推荐技术提供了一种新的方向。最后,本文结合作者参与的高教社基于本体、语义和语用的智能化教育资源平台项目,设计了一个基于用户情境的资源推荐系统模型,以此来说明该方法在实际中的应用。
With the rapid advancement of the Internet, the information overload has been a primal problem of Internet users. Resource recommendation system provides a very effective means of solving this problem. It can provide users with information filtering and the service of resource recommendation to improve their work efficiency, which is gradually welcomed by most users. So the research of resource recommendation is becoming a vital area. Presently, the main resource recommendation systems are system based on rules, one on contents and collaborative filtering recommendation.
System based on rules is formed from user and rule model. It can meet users’real-time needs, but the formulation of rules requires the involvement of experts, and the efficiency will be lowered, for the emergence of deviation and problems that is difficult to update. System based on contents chooses what is going to be recommended according to the contents of the information and the user interest. The shortcoming is that it relays only on user interest, and cannot find where the new interest of user. Therefore, its recommendation is restricted in the area of history list.
Among current resource recommendation systems, the collaborative filtering recommendation is the most successful one and extensively used in a number of resource recommendation systems. Collaborative filtering recommendation starts from user to seek the nearest neighbor user of the object user. By evaluating the other resource’s weighted evaluation value of the nearest neighbor as the target user’s expectation, it provided the resources to user, So the potential interest of user will be found. Though with a good many advantages, collaborative filtering recommendation has been successfully applied, the traditional collaborative filtering recommendation based on the project grade calculates similar neighbors of the object user then recommends resource. This will cause data sparsity and cold start, and the interest of users can not be fully reflected by the grade.
To settle problem above, the paper proposes a recommendation system based on pragmatics. On the basis of the traditional collaborative filtering recommendation, introduce the situation factor and factors of trust of pragmatics into the research of resource recommendation. Among all these information, the contextual information and scene information that influence user’s behaviors turn to user scenarios, and the user scenarios that militates user decision turns to the most prominent scenarios. This method acquires user’s prominent scenarios in the beginning, combines them to calculates user interest relativity, rating similarity and rating correlation matrix, then forms the correlation model of user interest, finally, calculates user confidence grade. Taking the user confidence grade as user score weight, combine the traditional method to predict the grade of object user.
This paper adopts Matlab and dose an simulation experiment in MovieLens. The experiment data shows that the method has a certain feasibility and effectiveness, and compared with the traditional one, the Slope One algorithm can reduce absolute average error. And it provides a new direction for current resource recommendation technology. Finally, combining with the intelligent educational platform project of Higher Education Press based on ontology, semantic and pragmatic, the paper designs a resource recommendation model based on user learning situation to explain the practical application of this method.
引文
[1]中国互联网络信息中心.第26次中国互联网络发展状况统计报告[R]. 2010年7月15日.4-5
[2]庞雅丽,王彩芬.个性化信息过滤技术[J]甘肃科技, 2007,(03)
[3] Singh, M.P., The pragmatic web. IEEE Internet computing, 2002: p. 4-5.
[4] Singh, M.P., The pragmatic web: preliminary thoughts. NSF-OntoWeb Workshop on Database and Information Systems Research for Semantic Web and Enterprises, Amicalola Falls State Park, GA, 2002.
[5] Sarwar, B., et al., Item-based collaborative filtering recommendation algorithms, in Proceedings of the 10th International Conference on World Wide Web. 2001: Hongkong. p. 285-295.
[6] Good, N., Schafer, J.B., Konstan, J.A., Borchers, A.,Sarwar,B.,Herlocker, J., Ricdl,J..Combining collaborative filtering with personal agents for better recommendations[C].In Proceeding of the 16th National Conference on Artificial Intelligence (AAAI-99),1999:439-446
[7]钟义信,面向智能研究的全信息理论:纪念Shannon信息论50周年.北京邮电大学学报, 1998. 21(004): p. 1-6.
[8] McKnight D.H, Chervany N L. The Meaning of Trust.. Technical Report MISRC Working Paper Series 96-04, University of Minnesota. Management Information Systems Research Center. 1996
[9] Sarwar, B.M., Konstank, J.A., Borchers, A.,Herlocker,J., Miller, B.,Riedl, J..Using filtering agents to improve prediction quality in the GroupLens research collaborative filtering system[C]. In Proceeding of CSCW’98, Scattle, WA,1998
[10] Co-funder.个性化推荐技术漫谈[N]. CSDDN特约专稿,2007.4.19
[11]赵庆峰.个性化信息服务系统理想模型的构建[J].情报杂志,2006,2:243-245
[12]张韧弦,形式语用学导论.复旦博学.语言学系列. 2008,复旦大学出版社:上海.
[13] Canhoto, A.I., Ontology-based interpretation and validation of mined knowledge: Normative and cognitive factors in data mining. Data mining with ontologies: implementations, findings, and frameworks. 2007: Hershey, NY, Information Science Reference. 84-105.
[14] Ansari, A., S. Essegaier, and R. Kohli, Internet recommendation systems. Journal of Marketing Research,2000.37(3):p.363-375.
[15]邓爱林,朱杨勇,施伯乐.基于项目评分预测的协同过滤推荐算法[J].软件学报.2003.Vo14No,9.
[16] Sarwar, B.M., Karypis, G, Konstan, J.A., Riedl, J..Analysis of recommendation algorithms for E-commerce[C]. In Proceedings of the ACM EC’00 Conference. Minneapolis, MN,2000:158-167
[17] Morris C. Foundations of the Theory of Signs.Chicago:University of Chicago Press,1938
[18] Morris C.Signs,Language and Behavior.Englewood cliffs:Prentice Hall,1946
[19]何自然.我国近年来的语用学研究.现代外语,1994,(4):13~17
[20]刘根辉.从语用学发展历程看中国语用学研究的发展方向.外语学刊,2005,(1):18~23
[21]钱冠连.论构建语用推理模式的出发点——新格赖斯理论评论.现代外语,1994,(3):1~6
[22]刘根辉,计算语用学基础理论及其应用研究,2005
[23] Stephen Weeks.Understanding trust manage-ment systems In:Proc.of the 2001 IEEE Symposium on Security and Privacy,Oakland,California.IEEEComputerSociety,2001:94-105
[24] Gazdar G.Pragmatics:Implicature,Presupposition and Logical Form.New York:
[25] Wilks Y.A.Preference semantics.In E.Keenan ed.The Formal Semantics of NaturalLanguage.Cambridge:Cambridge University Press,1973
[26] Montague R.Pragmatics.In:R.Klibanskyed.Contemporary Plulosophy:A Survey Florence:La Nuova Italia Editrice,1968.102-122.Reprinted in:R.H.Thomasoned.Formal Plulosophy:Selected Papers of Richard Montague.New Haven and London:Yale University Press,1974.95-118.
[27] Deemter K. van, G. Brockhoff, H. Bunt et. From TENDUM to SPICOS, or: How flexible is the TENDUM approach to question answering? In:IPO Annual Progress Report 20,1985.83-90.
[28]宁宇鹏,曹之大。开放网络中信任模型研究.计算机工程.2005,31(4):144-145,171.
[29]刘玉龙,曹元大.分布式环境主观信任模型研究.北京理工大学学报,2005.25(6);504-508
[30] Lemire, D. and A. Maclachlan. Slope One predictors for online rating-based collaborative filtering. 2005: Society for Industrial Mathematics.
[31] Fodor J.A.The modularity of mind.Cambridge,MA:MIT Press,1983.
[32]姚忠,吴跃,常娜,集成项目类别与语境信息的协同过滤推荐算法[J].计算机集成制造系统, 2008. 14(007):1449-1456.
[33]李聪,梁昌勇,马丽,基于领域最近邻的协同过滤推荐算法[J].计算机研究与发展, 2008. 45(009):1532-1538.
[34]王宏宇,商务推荐系统的设计研究[D]. 2007,中国科技大学.
[35] Herlocker, J.L., et al., Evaluating collaborative filtering recommender systems [J]. ACM Transactions on Information Systems, 2004. 22(1): 5-53.
[36] Turing A.M.Computing machinery and intelligence.Mind,1950,59:433~460.
[37] Liang, T.-P., et al., A semantic-expansion approach to personalized knowledge recommendation.Decision Support Systems, 2007. In Press, Corrected Proof.
[38]高旻,基于计算语用学和项目的资源协同过滤推荐研究[D].重庆大学博士论文.2010.
[39] Sun, L. and K. Ousmanou, Articulation of information requirements for personalised knowledge construction [J]. Requirements Engineering, 2006. 11(4): 279-293.
[2]庞雅丽,王彩芬.个性化信息过滤技术[J]甘肃科技, 2007,(03)
[3] Singh, M.P., The pragmatic web. IEEE Internet computing, 2002: p. 4-5.
[4] Singh, M.P., The pragmatic web: preliminary thoughts. NSF-OntoWeb Workshop on Database and Information Systems Research for Semantic Web and Enterprises, Amicalola Falls State Park, GA, 2002.
[5] Sarwar, B., et al., Item-based collaborative filtering recommendation algorithms, in Proceedings of the 10th International Conference on World Wide Web. 2001: Hongkong. p. 285-295.
[6] Good, N., Schafer, J.B., Konstan, J.A., Borchers, A.,Sarwar,B.,Herlocker, J., Ricdl,J..Combining collaborative filtering with personal agents for better recommendations[C].In Proceeding of the 16th National Conference on Artificial Intelligence (AAAI-99),1999:439-446
[7]钟义信,面向智能研究的全信息理论:纪念Shannon信息论50周年.北京邮电大学学报, 1998. 21(004): p. 1-6.
[8] McKnight D.H, Chervany N L. The Meaning of Trust.. Technical Report MISRC Working Paper Series 96-04, University of Minnesota. Management Information Systems Research Center. 1996
[9] Sarwar, B.M., Konstank, J.A., Borchers, A.,Herlocker,J., Miller, B.,Riedl, J..Using filtering agents to improve prediction quality in the GroupLens research collaborative filtering system[C]. In Proceeding of CSCW’98, Scattle, WA,1998
[10] Co-funder.个性化推荐技术漫谈[N]. CSDDN特约专稿,2007.4.19
[11]赵庆峰.个性化信息服务系统理想模型的构建[J].情报杂志,2006,2:243-245
[12]张韧弦,形式语用学导论.复旦博学.语言学系列. 2008,复旦大学出版社:上海.
[13] Canhoto, A.I., Ontology-based interpretation and validation of mined knowledge: Normative and cognitive factors in data mining. Data mining with ontologies: implementations, findings, and frameworks. 2007: Hershey, NY, Information Science Reference. 84-105.
[14] Ansari, A., S. Essegaier, and R. Kohli, Internet recommendation systems. Journal of Marketing Research,2000.37(3):p.363-375.
[15]邓爱林,朱杨勇,施伯乐.基于项目评分预测的协同过滤推荐算法[J].软件学报.2003.Vo14No,9.
[16] Sarwar, B.M., Karypis, G, Konstan, J.A., Riedl, J..Analysis of recommendation algorithms for E-commerce[C]. In Proceedings of the ACM EC’00 Conference. Minneapolis, MN,2000:158-167
[17] Morris C. Foundations of the Theory of Signs.Chicago:University of Chicago Press,1938
[18] Morris C.Signs,Language and Behavior.Englewood cliffs:Prentice Hall,1946
[19]何自然.我国近年来的语用学研究.现代外语,1994,(4):13~17
[20]刘根辉.从语用学发展历程看中国语用学研究的发展方向.外语学刊,2005,(1):18~23
[21]钱冠连.论构建语用推理模式的出发点——新格赖斯理论评论.现代外语,1994,(3):1~6
[22]刘根辉,计算语用学基础理论及其应用研究,2005
[23] Stephen Weeks.Understanding trust manage-ment systems In:Proc.of the 2001 IEEE Symposium on Security and Privacy,Oakland,California.IEEEComputerSociety,2001:94-105
[24] Gazdar G.Pragmatics:Implicature,Presupposition and Logical Form.New York:
[25] Wilks Y.A.Preference semantics.In E.Keenan ed.The Formal Semantics of NaturalLanguage.Cambridge:Cambridge University Press,1973
[26] Montague R.Pragmatics.In:R.Klibanskyed.Contemporary Plulosophy:A Survey Florence:La Nuova Italia Editrice,1968.102-122.Reprinted in:R.H.Thomasoned.Formal Plulosophy:Selected Papers of Richard Montague.New Haven and London:Yale University Press,1974.95-118.
[27] Deemter K. van, G. Brockhoff, H. Bunt et. From TENDUM to SPICOS, or: How flexible is the TENDUM approach to question answering? In:IPO Annual Progress Report 20,1985.83-90.
[28]宁宇鹏,曹之大。开放网络中信任模型研究.计算机工程.2005,31(4):144-145,171.
[29]刘玉龙,曹元大.分布式环境主观信任模型研究.北京理工大学学报,2005.25(6);504-508
[30] Lemire, D. and A. Maclachlan. Slope One predictors for online rating-based collaborative filtering. 2005: Society for Industrial Mathematics.
[31] Fodor J.A.The modularity of mind.Cambridge,MA:MIT Press,1983.
[32]姚忠,吴跃,常娜,集成项目类别与语境信息的协同过滤推荐算法[J].计算机集成制造系统, 2008. 14(007):1449-1456.
[33]李聪,梁昌勇,马丽,基于领域最近邻的协同过滤推荐算法[J].计算机研究与发展, 2008. 45(009):1532-1538.
[34]王宏宇,商务推荐系统的设计研究[D]. 2007,中国科技大学.
[35] Herlocker, J.L., et al., Evaluating collaborative filtering recommender systems [J]. ACM Transactions on Information Systems, 2004. 22(1): 5-53.
[36] Turing A.M.Computing machinery and intelligence.Mind,1950,59:433~460.
[37] Liang, T.-P., et al., A semantic-expansion approach to personalized knowledge recommendation.Decision Support Systems, 2007. In Press, Corrected Proof.
[38]高旻,基于计算语用学和项目的资源协同过滤推荐研究[D].重庆大学博士论文.2010.
[39] Sun, L. and K. Ousmanou, Articulation of information requirements for personalised knowledge construction [J]. Requirements Engineering, 2006. 11(4): 279-293.