摘要
为了实现机器人理解自然语言的目的,提出了一种应用于中文服务指令的自主处理方法,旨在将中文服务指令直接映射为动作序列.首先深入地分析收集的中文服务指令,提炼出了关键信息与句法结构的对应关系.在此基础上,以短语组块的语法规则为基础,利用概率模型提取指令中的关键信息.为了解决指令解析问题,设计了一种服务任务分解模板,机器人依据该模板可以将中文服务指令直接映射为动作序列.最后进行仿真实验,统计了中文服务指令自主处理过程中各个环节的实验结果,其中组块标注、信息提取及指令解析的准确率平均值分别达到了92.9%、92.7%和97.2%,验证了本文方法的有效性和可行性.
An autonomous processing method of Chinese service instruction is investigated to make robot understand natural language. This method directly maps Chinese service instructions to sequences of executable actions. Firstly, the collected corpus of the service instruction expressed in Chinese is deeply studied, and the corresponding relationship between the key information and syntactical structure is presented. According to the grammar rules of phrase chunk, a probability model is used for key information extraction. In order to resolve the problem of task decomposition, a task decomposition template （TDT） is presented. Based on this, the service instruction is mapped to sequences of actions. At last, the simulation about autonomous Chinese instruction processing is performed, and the experimental results of every step are provided, with the average accuracies of 92.9% in chunk marking, 92.7% in information extraction, and 97.2% in instruction parsing, which validates the correctness and availability of the proposed method.
出处
《机器人》
EI
CSCD
北大核心
2015年第4期424-434,共11页
Robot
基金
国家自然科学基金项目(61305113)
河北省自然科学基金项目(F2012203199)
关键词
中文服务指令
句法分析
信息提取
指令解析
Chinese service instruction
syntactic analysis
information extraction
instruction parsing
