| 语义脚色标注是今朝语义剖析的一种重要完成措施,它也是最近几年来天然说话处置范畴的一个研究热门,是信息抽取、信息检索、浏览懂得、问答体系等多种天然说话处置技巧的主要基本。今朝,语义脚色标注的研究才方才起步,对语义脚色标注的研究尚少,其研究程度也有待进一步的进步。本文研究的语义是基于山西年夜学开辟的汉语框架语义网。本文重要完成了以下几方面任务1、汉语框架语义网简介。汉语框架语义网是以Fillmore的框架语义学为实际基本,参照FrameNet,以汉语真实语料为根据而构建。汉语框架语义网中界说的框架元素就是平日所说的语义脚色,汉语框架语义网为语义脚色标注供给了异常好的语料资本。。2、模仿英语框架元素主动标注的办法停止汉语框架元素主动标注。英语框架元素标注是先辈行句法剖析,把句法信息作为特点停止框架元素主动标注。是以,本文对汉语句法剖析和浅层句法剖析停止比拟,肯定在浅层句法剖析基本长进行框架元素主动标注。3、提出汉语框架语义主动标注的办法和步调。本文提出了一种基于前提随机场模子停止框架语义主动标注办法。该办法以词作为标注的根本单位,在低层前提随机场模子中处理了框架元素辨认,将辨认成果传递到下层短语类型辨认的前提随机场模子,然后再将辨认成果传递到下层句法功效辨认的前提随机场模子。选用汉语框架语义网中“陈说”、“包括”、“具有”、“文本”、“艺术成品”框架句子库中的句子作为语料,用上述办法完成了这些框架句子库中句子的主动标注,其框架元素主动标注成果的召回率58%-75.7%,精确率66.1%—83.9%,F值61.8-79.6%。最初,对本课题研究停止了总结剖析,指出个中存在的缺乏。 Abstract: Semantic role labeling is the current semantic analysis is an important method to finish. It is also in recent years to the natural language processing field a research hotspot is information extraction, information retrieval, know how to browse, question answering system, such as a variety of natural language disposal techniques mainly basic. Today, semantic role labeling research has just started, research on semantic role labeling is less, the research level is also needs to be further progress. This paper is a study of the semantics of Chinese FrameNet based on the Open University of Shanxi. This paper completed the following tasks, 1 Chinese framenet. Chinese FrameNet is based on frame semantics for Fillmore basic theory, according to FrameNet, the Chinese corpus is constructed according to the. Definition in Chinese FrameNet frame elements is semantic role normally say, Chinese FrameNet semantic role annotation provides exceptionally good corpus capital. 2, the imitation of English active annotation frame elements to stop Chinese frame elements active annotation. English annotation is the first frame element analysis of syntax, the syntactic information as the characteristics of frame elements of active annotation. This paper is to analysis on Chinese syntactic parsing and shallow syntactic comparison, certainly in the shallow syntactic analysis of the basic framework of long active annotation elements. 3, put forward the Chinese framework semantic annotation method and the step of active. This paper presents a random field model of active annotation approach based on the premise of semantic framework. The approach to word as marked the fundamental unit, in low layer premise airport model with processing the frame element identification, the recognition results transmitted to the lower phrase types to identify the premise of random field model. And then, the recognition results transmitted to the lower syntactic functions to identify the conditional random field model. Sentences selected Chinese FrameNet "Chan", "including", "with", "text",, "art of the finished product" sentence structure in the library as the corpus, used in this way, complete the sentences in the sentence structure library of active annotation, the frame element active annotation results recall 58%-75.7%, accurate rate 66.1%, 83.9%, the value of F 61.8-79.6%. At first, the article summarized the analysis, pointed out the existing deficiency. 目录: 中文摘要 4-5 ABSTRACT 5-6 第一章 引言 11-15 1.1 探讨的背景和意义 11 1.2 国内外探讨近况 11-13 1.3 论文的组织安排 13-15 第二章 汉语框架语义网简介 15-23 2.1 汉语框架语义网 15-19 2.1.1 框架库 15-17 2.1.2 句子库 17-19 2.1.3 词元库 19 2.2 汉语框架语义网运用举例 19-23 第三章 句法略论 23-37 3.1 句法略论性能比较 23-30 3.2 浅层句法略论 30-37 3.2.1 条件随机场(CRF)的定义 30-32 3.2.2 特征模板设置 32-33 3.2.3 “BIO”策略 33-34 3.2.4 浅层句法略论实验 34-37 第四章 汉语框架语义自动标注 37-51 4.1 汉语框架语义自动标注 37-43 4.1.1 层叠条件随机场 37-38 4.1.2 “陈述”“包含”框架句子的标注实验 38-42 4.1.3 语义自动标注的步骤 42-43 4.2 其它框架语义标注实验结果 43-46 4.2.1 “拥有”框架语义标注实验结果 43-44 4.2.2 “文本”框架语义标注实验结果 44-45 4.2.3 “艺术制品”框架语义标注实验结果 45-46 4.3 旅游问答系统样本库的框架语义自动标注 46-48 4.4 实验结果比较略论 48-51 第五章 结束语 51-52 附录 52-54 参考文献 54-56 发表文章目录 56 个人简历 56-57 致谢 57-58 |
