自然语言理解的全信息方法论  

Comprehensive Information Based Methodology for Natural Language Understanding

作　　者：钟义信[1] 

机构地区：[1]北京邮电大学智能科学技术研究中心,北京100876

出　　处：《北京邮电大学学报》2004年第4期1-12,共12页Journal of Beijing University of Posts and Telecommunications

基　　金：国家自然科学基金项目(69982001);国家"863计划"项目(2001AA114210)

摘　　要：在经济全球化需求的推动下,世界在酝酿一场"自然语言信息技术革命",它的基础和核心是"自然语言理解"的理论与方法."全信息自然语言理解方法论"是作者的"全信息理论"在自然语言理解领域的应用.与文献中已有的其他工作不同,其主要特色是:一方面,试图实现语法信息、语义信息、语用信息的综合利用;另一方面寻求"规则方法"和"统计方法"的和谐互补,从而有效增强对自然语言的理解能力.近几年来,应用这一方法论完成了一系列自然语言理解方面的课题,取得了一批可喜的研究成果,表明全信息自然语言理解方法论具有很好的前景.The Comprehensive Information Based Methodology for Natural Language Understanding presented in the paper is an application of the Comprehensive Information (CI) Theory, which was established by the author of the paper years ago for meeting the needs in artificial intelligence research. Different from other works in the field of natural language processing, the major features of the methodology presented in the paper include two folds. Firstly, instead of the syntactic or/and semantic information is used, all components of the comprehensive information are fully utilized for natural language understanding. Secondly, instead of using one or another approach preferably, both the rule-based approach and the statistics-based approach have been complementarily integrated. The successes of the research projects guided by the methodology and completed last few years prove the methodology promising.

关 键 词：自然语言理解 全信息理论 规则方法 统计方法 全信息自然语言理解方法论 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]