基于序列标注算法的命令语义理解.docxVIP

42 - 哈尔滨工业大学本科毕业设计（论文） 基于序列标注算法的命令语义理解 摘 要 人工智能大热，智能聊天机器人成为研究热门，语义理解是其中至关重要的一部分。本文研究基于序列标注算法的命令语义理解，讲述了订票命令的语义理解和检索过程。 本文研究的指令范围限定在订机票和订火车票领域。我们通过条件随机场模型来进行序列标注，从而得到输入命令中与订票有关的信息—我们称之为槽。得到订票所必需的槽（出发日期，出发城市，到达城市）后，即可进行相应的查询，并将查询到的结果返回给用户。 本文分为五章。第一章介绍了课题的背景意义以及国内外的研究现状，具体阐述了我们的研究内容。第二章至第四章阐述了各个模块的工作：第二章讲了数据的来源和标注方法，给出了设计的槽类型标签及释义。第三章讲述了序列标注模块中所做的工作。第四章讲述了在通过以上两步获得了槽信息后，系统如何通过这些槽的内容检索到正确的信息并输出给用户。 我们采用PRF评价体系来评估模型效果。目前训练的CRF模型整体P,R,F值均达到了95%以上。输入命令后，系统能理解大多数情况下的订票语义，并返回正确的信息。具体的结果分析和示例在第五章给出。 关键词：CRF；序列标注；语义理解 Abstract With the popularity of Artificial intelligence, chat robot becomes an active research area. semantic understanding is one of most important part of this. This paper focuses on order semantic understanding based on sequence labeling algorithms. Instruction scope of this study is limited to the field of book flights and train tickets. We use CRFs model for sequence labeling , thus to obtain information related with the booking - which we call slots . After obtaining all the necessary slots(start date, start city, arrival city), the system we designed can make the appropriate query, and the query results will be returned to the user. This paper is divided into five chapters. The first chapter introduces the background , significance of the research ,and research status at home and abroad. Chapter 2 concerns the source of the data set and labeling methods, the design and interpretation of the slot type labels. The third chapter describes what the sequence labeling part does. The fourth chapter describes that after the above two steps , how to retrieve content to get the right information and output it to the user. We use PRF evaluation system to evaluate the model results . The overall P, R, F value of our current model of CRFs reached more than 95 percent . Enter the command, the system can understand the semantics of tickets in most cases, and returns the correct information . Specific analysis of the results and examples are given in the f