面向文本标识柔性语义相似性度量方法-软件工程专业毕业论文.docx

目 录第一章绪论······························· 课题背景 ····························· 国内外现状 ···························· 语义相似性度量 ······················· 新闻故事分割技术 ······················ 新闻故事协同分割技术 ··················· 术语和标记说明 ························· 本文贡献 ····························· 论文结构 ····························· 8第二章 柔性语义相似性度量······················ 语义相关性 ···························· 上下文相关性························· 故事内相关性························· 故事间相关性························· 共字相关性·························· 相关性关系图··························· 相关性传播算法 ························· 词语级别柔性语义相似性度量·················· 句子级别柔性语义相似性度量·················· 16第三章 基于柔性语义相似性的新闻故事分割············· 图模型的建立··························· 归一化准则 ···························· 动态优化解法··························· 20第四章 基于柔性语义相似性的新闻故事协同分割 ·········· 初始化······························· 前景/背景故事建模························ 构建用于新闻故事协同分割的吉布斯能量方程 ········ 能量方程优化和模型修正 ···················· 25第五章 实验结果与分析 ························ 在新闻故事分割中的实验 ···················· 实验文集以及实验设定 ··················· 与现有新闻故事分割方法的比较 ·············· 与现有语义相似性度量方法的比较············· 在新闻故事协同分割中的实验·················· 实验文集以及实验设定 ··················· 与现有的新闻故事协同分割算法的比较 ·········· 与现有的语义相似性度量方法的比较············ 38第六章 总结与展望 ··························· 总结································ 展望································ 45参考文献 ································· 46发表论文和参加科研情况说明 ····················· 49致 谢··································· 50第一章 课题背景随着信息技术的的普及和发展,例如广播新闻、会议记录、网上公开课之类的多媒体内容正在互联网上急速增加,人们亟需一种快速有效的方法将这类多媒体数据进行自动的组织和管理,以用于基于主题的文本检索和分析。一个多媒体文档,例如一小时的广播新闻节目,通常由多个故事组成,为了进行高效率的语义检索,指导使用者找到他们感兴趣的主题的开始和结束是很重要的,同时,一个分割好的多媒体文档也是进行主题跟踪[1]、语义总结[2]、信息检索[3]、用户行为分析[4][5]等高层次的语义浏览的重要前提条件。新闻故事分割技术[6][7][8]和协同分割技术[9]的目的就在于将新闻故事脚本分割成主题一致的故事。从技术上讲,新闻故事分割技