面向领域自然语言处理方法的研究-计算机技术专业毕业论文.docxVIP

上海交通 上海交通大学工程硕士学位论文 面向领域自然语言处理方法的研究 摘要 自然语言处理的目标是让人可以使用自然语言与计算机进行通讯，是人工智 能的重要方向之一。自然语言其本身固有的复杂性和不确定性是任何自然语言处 理方案所要面对和解决的难题。本文绕开了通用自然语言的复杂性而专注于研究 特定领域的自然语言，通过对面向领域自然语言、面向领域自然语言的处理以及 相关处理方法的分析，提出了一种面向领域的自然语言处理方法。允许在特定的 语境下，借助预先定义的面向特定领域的领域词典以及面向特定领域的规则库完 成对于输入的自然语言文本的解析，从而实现特定领域下对文本携带语义信息的 理解。 在一些需要进行特定目的的人机交互场合下可以发挥巨大的作用。论文 分析了多个自然语言处理应用的特点和模式 ，提出了一个通用的自然语言处理 方案以处理自然语言文本挖掘和 DSNL 两种标准的自然语言处理应用。方案在 传统的词法分析，语法解析间加入了概念对象解析步骤以更好的适应自然语言内 在的不确定性特性。为了方便的建立领域规则库和领域词典，本文还提出了一种 自动挖掘面向领域的语法规则的算法,使用二阶段算法从语料中反向推导规则。 论文设计了一个对比实验证明了面向领域自然语言相比传统方式有明显的性能 提升。 关键字 面向领域,自然语言处理,人工智能 RESEARCH FOR DOMAIN SPECIFIC NATURAL LANGUAGE PROCESSING SOLUTION ABSTRACT Natural Language Processing (aka NLP) which is one of the most important branches in AI researches, is for the purpose of enabling computer to understand natural language from a natural man. Any solution that attempts to go through NLP would have to go over the hard obstacle: the inherent intricateness and arbitrariness of natural language .This paper suggests a domain specific NLP solution to go around these obstacles by focusing on several specific scenarios with limited domain dictionary and language grammar rules and provide a way for computer to understand semantic information by parsing input text. This paper also describes the basic flows and related concepts of this solution in details. This solution is ideal for the cases where involve large amount of interaction between human and computer. This paper looks deep into NLP applications from many different domains and analysis shared patterns among them. The NLP solution suggested in this pager is supposed to serve as a common purpose solution for two typical domain specific NLP pattern: natural language text mining and DSNL. The solution adds a “concept cognize” phase between “lexical analyze” phase and “syntax parsing” phase for better adoption with the inherent uncertainties of natural language. And for ease the creation of domain base, this paper discussed an algorithm for min