多智能体并行设计环境中的自然语言通信功能

The natural language communication function in a multi-agent parallel design environment

研究了一个多智能体并行设计环境中自然语言通信模块的设计和实现方法 .该设计方法建立在受限语言的基础上 ,突出了在协商式多智能体之间进行自然语言通信时的准确性 .通过建立一个受限词汇的综合信息库 ,记录词汇的各种语法、语义信息 ;描述了词库的结构和形态 ;为了区分基本词汇和面向用户领域词汇 ,将词库分级 ;在词法分析时采用总库索引来查询词汇库 ,并提出了提高词库性能的方法 ;限定了用户使用的基本句型 ,采用扩充短语结构文法对句子结构进行分析 ,并产生相应的语义逻辑公式 ;定义了句子生成模板 ,智能体借助该模板提供的接口函数生成自然语言句子 .此外 ,还提出了提高通信有效性的方法 ,为智能体提供了灵活、友善的接口 ,提高了多智能体并行设计环境的性能 .

The design methods and implementation techniques of a natural language communications module in a multi agent parallel design environment were described. The method is based on the restricted language and gives prominence to the veracity of the communications between the negotiation′s multi agents. The authors design a restricted English subset, build a dictionary that includes the restricted words′ grammatical and semantic attributes, build an index for the dictionary, and look up words through it in the course of morphology analyzing, at the same time we give some methods to improve the performance of the dictionary, restrict the sentence styles used by users and make use of augmented phrase structure grammar, put forward some methods to improve the communications veracity in the restricted language environment. The agents can call functions to make natural language sentences based on templates, provides an agile, friendly interface and improves the performance of the environment.