当代人工智能表征的分解方法及其问题  

The Analysis Method of Contemporary Artificial Intelligence Representation and Its Problems

作　　者：殷杰[1] 董佳蓉[1] 

机构地区：[1]山西大学科学技术哲学研究中心,山西太原030006

出　　处：《科学技术与辩证法》2009年第2期23-28,89,共7页Science Technology and Dialectics

基　　金：霍英东教育基金会高等院校青年教师基金项目(111105);全国优秀博士学位论文作者专项资金资助项目(200503)

摘　　要：分解方法是建立在形式系统之上的人工智能表征的必然选择,也是造成现阶段人工智能表征在自然语言语义理解方面各种瓶颈问题的理论根源。文章认为,句子层次结构以及三个平面的划界理论是分解方法难以实现段落或篇章语义理解的关键所在。基于词汇的语境描写方法难以突破单句限制,人工智能表征要想获得突破,就必须借助基于段落或篇章的整体性语境描写方法,而这正是分解方法所缺失的。多年来,人工智能表征取得的成就表明,脱离分解方法去谈整体性语境构建方法是不切实际的。整体性语境构建方法应当以分解方法为基础,二者之间是一脉相承而非矛盾的关系。二者的有机融合,是解决人工智能表征分解方法瓶颈的关键所在。The analysis method is the necessary selection of artificial intelligence representation which is based on formal system.It is also the theoretical root of various bottleneck problems of artificial intelligence representation in the natural language semantic understanding at present.Therefore,it is necessary to reflect on it.The hierarchical structure of sentences and the demarcation of the three-plane theory are the crux for the analysis method not to be able to achieve the semantic understanding of paragraph or text.The context description method which is based on the word is difficult to break through the restriction of the single-sentence.In order to get a breakthrough,the artificial intelligence representation must use the holistic context description method which is based on the paragraph or the text.And this is what the analysis method is missing.Over the years,the achievements of artificial intelligence representation showed that it is unrealistic to talk about the holistic context constructing method without the analysis method.The holistic context constructing method should be based on the analysis method,and they are not contradictory.The combination of the two is the key to solving the bottlenecks of the analysis method of artificial intelligence representation.

关 键 词：人工智能表征 分解方法 整体性语境构建方法 

分 类 号：N02[自然科学总论—科学技术哲学]