基于XML的知识表示体系和汉语口语对话系统通用模型研究
文献类型：学位论文
| 作者 | 汪志鸿 |
| 学位类别 | 博士 |
| 答辩日期 | 2005 |
| 授予单位 | 中国科学院声学研究所 |
| 授予地点 | 中国科学院声学研究所 |
| 关键词 | 口语对话系统 知识表示 语音识别 自然语言理解 |
| 其他题名 | XML Based Knowledge Representation Architecture and General Model for Chinese Spoken Dialogue Systems |
| 中文摘要 | 口语对话系统是当前人机交互领域的一个重要的研究课题。本文研究如何建立对话系统中统一的知识表示模型和如何建立通用的口语对话系统的方法，使口语对话系统的开发更加规范化、工程化和实用化。论文的主要贡献如下：（1）建立了使用XML语言描述的系统知识表示体系标准SDSXML。在系统研究知识表示方法的基础上卜将口语对话系统的知识表示体系分为基本集、数据集和规则集三个部分，用基本集描述口语基本的语法，语义和语用特征，用数据集描述系统能够提供给用户的信息，用规则集描述根据用户话语的语用语义信息进行对话控制的规则。SDSXML为面向任务的口语对话系统提供了统一的知识表示体系和接口，便于口语对话系统的数据共享和软件复用。（2）建立了口语对话系统通用模型，并在此模型的基础上搭建了一个语音银行VoiceBank实验系统，验证了SDSXML在知识表示方面的通用户性和系统各个组成模块的独立性，提高了通用模块的可移植性。（3）提出了通过系统期待建立语言模型来指导语音识别的方法。通过系统期待的指导，结合对话系统的当前状态，建立涵盖用户所有可能应答的用SDSXML表示的对话规则的集合，然后根据这个规则集合设置识别网络，降低语音识别器的搜索空间，提高识别的效率和准确率。 |
| 英文摘要 | Spoken Dialogue System has been one of the important research subjects in human-computer interaction area in recent years. This thesis is focused on building the uniform knowledge representation model in the dialogue system and the general approach method of spoken dialogue system. We aim at the standardization of the dialogue system more. The research work will make the development process more practical, and realize the engineering development methods. The main contributions of this thesis are: 1. Building a system knowledge representation architecture standard SDSXML described with XML in spoken dialogue system. The thesis investigates the knowledge representation methods, and divides the architecture into three part: basic aggregation which describe the syntax, semantics and pragmatics of the spoken language, data aggregation which describe the information provided to the user, and rule aggregation which describe the dialogue control rules according to the user utterance. Since SDSXML provides the uniform knowledge representation architecture and the uniform interface for the task-oriented spoken dialogue systems, data sharing and software reuse will be possible. Building the general model of spoken dialogue system, and realize the VoiceBank system based on it. VoiceBank verifies the generality of SDSXML in the aspect of knowledge representation. It's modules are independent of each other, and easy to reuse in other systems. The thesis present a speech recognition method directed by system expectation. The collection of dialogue rules represented by SDSXML has been built directed by system expectation with the system current state. Then the recognition network is setting according to the collection in order to reduce the search space of the speech recognizer and improve the recognition accuracy. The experiment results is satisfied. |
| 语种 | 中文 |
| 公开日期 | 2011-05-07 |
| 页码 | 95 |
| 源URL | [http://159.226.59.140/handle/311008/1038]  |
| 专题 | 声学研究所_声学所博硕士学位论文_1981-2009博硕士学位论文 |
| 推荐引用方式 GB/T 7714 | 汪志鸿. 基于XML的知识表示体系和汉语口语对话系统通用模型研究[D]. 中国科学院声学研究所. 中国科学院声学研究所. 2005. |
入库方式： OAI收割
来源：声学研究所
浏览0
下载0
收藏0
其他版本
除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。