【作者】 王海燕；

【导师】 蒋剑平；

【作者基本信息】 大连海事大学 ， 计算机应用技术， 2008， 硕士

【摘要】 网络技术的出现和发展使得基于网络的远程教育作为一种新的教学模式迅速发展起来。它克服了传统教育在时间、空间上的限制,提供了便捷的学习方式、多样的学习环境和丰富的学习资源。然而这种教育模式将教师和学生分离开来,无法进行面对面的交流与解释,学习者的疑问无法得到及时的解答,因此建立智能答疑系统模型就成了开发网络教学系统的重要任务之一。答疑系统作为网络教育的一个重要组成部分,在帮助学生解除疑惑、获取知识、加强师生交流方面发挥着重要作用。然而传统的答疑系统主要是基于关键词匹配,不具有语义理解功能,因而不能实现问题答案的语义级的匹配,远远不能满足日益增长的网络教育需要。本文在深入研究自动问答相关理论和技术的基础上,分析了智能答疑的关键技术,针对用户问句分词不准确的问题,采用了基于概念词典和常用词典的字符串逆向最大匹配分词算法,提高了分词的准确度;针对标注后的问句向量与常问问题集中问题匹配度比较低的问题,采用了关键词的规范化机制,从而使问句尽量与问题集中问题匹配,省略了后面复杂的相似度计算,提高检索速度;针对问句关键词扩展后“噪音”信息过多的问题,建立了关键词的分级扩展机制,避免了扩展后检索主题的漂移并给用户提示作用以启发式查找;针对传统的向量空间模型未考虑不同关键词在检索答案时影响不同的问题,提出了一种基于分解的向量空间模型和语义概念的问句相似度计算方法,提高了相似度计算的准确性。基于以上的关键技术,本文提出了一种新的网络智能答疑系统模型,并分析了模型中的各个处理模块,给出了系统模型的处理流程。本系统模型实现了答疑系统的语义理解,提高了现有网络答疑系统的效率,并具有一定的智能性。更多还原

【Abstract】 With the appearance and development of network technology,Distance Education based on network education develops rapidly as a new teaching mode.It breaks the time and space limitation and provides convenient learning ways and diversiform learning environment.But in this education mode,teachers and students are separated from each other and can not communicate face-to-face and the students’ problems can not be answered in time.So establishing an intelligent Question Answer System Model is one of important tasks of development of network education system.As one important part of network education,Question Answer System plays an important role at helping students rescind their doubts,knowledge acquisition and enhancing mutual interaction between teachers and students.However,the traditional Question Answer System cannot realize semantic matching of questions and answers and cannot not fill the require of increasing network educational needs because it is based on keyword matching mostly and short of semantic understanding.On the basis of research on the theories and technologies related to automatic Question Answering,the thesis analyzes the key technologies of Network Intelligent Question Answer.To solve the problem of segmentation inaccuracy of user question sentences,the paper puts forward the reverse maximum matching word segmentation algorithm based on concept dictionary and common dictionary improving the accuracy of segmentation.In view of the matching degree of question vector after marking and question of Frequently Asked Questions,the paper proposes a mechanism of keywords standardization making the matching degree as high as possible,avoiding the latter complex similarity computation and increasing the Retrieval speed.According to the problem of excessive noise information after keywords extension,the thesis proposes a mechanism of classified keywords expansion avoiding drift of retrieval theme and giving a hint to user for heuristic search.Aiming at different effects of keywords in retrieval are not considered in the traditional vector space model,the paper presents a question similarity computation approach based on splited vector space model and semantic concepts raising the accuracy of similarity computation.Based on the above key techniques,the article proposes a new Network Intelligent Question Answer System Model and analyzes each packet-handling module and gives the processing flow of the system model.The system model realizes semantic understanding of Question Answer system improving efficiency of the present Question Answer system and with certain intelligence.更多还原