项目名称: 面向事件分析的信息意图检测、建模与群体意图推理技术研究
项目编号: No.61462073
项目类型: 地区科学基金项目
立项/批准年度: 2015
项目学科: 其他
项目作者: 过弋
作者单位: 石河子大学
项目金额: 46万元
中文摘要: 近年来针对文本内容的事件研究工作持续向多样化发展(例如:事件抽取、新事件或热点事件检测、多层次分类、倾向性分析、信息结构分析、角色填充、事件指代消解、因果分析等),由于海量文本中的事件大幅增加,研究事件间关联关系或发展趋势的需求日益突出。事件之间并非相互孤立,文本内容中的信息意图或主观倾向性信息,对于揭示和发现事件间的大量隐性关联关系,以及掌握事件的发生和发展的规律(包括特定事件的起因、过程和结果)具有深远的研究意义。本项目将重点研究信息意图的语言学规律,从文本内容分析(语篇标记词、句法结构、意图性动词及句式等方面)着手,结合语篇所包含的时间/空间信息、逻辑语义关系和主位结构理论,注重语篇的整体信息结构,实现语篇内多层次信息意图自动检测、个体信息意图建模和群体意图推理,获取能够直接揭示事件间多种关联关系的意图关系集合,为研究事件关系推理和趋势预测开辟新的途径,丰富语篇层面的文本挖掘研究工作
中文关键词: 信息抽取;文本挖掘;知识发现;自然语言理解;自然语言处理
英文摘要: In recent years, research work for the textual event analysis has developed into diverse ways (e.g. event extraction, new and hot event detection, multiple layer classification, sentiment analysis, information structure analysis, cross-event inference, event pronoun resolution, and causality analysis). Due to the significant increase of events in massive texts, the research requirement for event correlations becomes increasingly prominent. Since events in texts are not isolated at all, informative intention or subjective tendency information are critical and significant to reveal and discover massive hidden or implicit connections among events and catch the occurrence and development pattern (including the cause, process and results of particular events). This project will make effort to research the linguistic theories about informative intention and the methodology of informative intention extraction at discourse level. The practical technologies include extracting discourse markers, syntactic structures, cognition verbs and relevant spatial/temporal information, solving logico-semantic problems, constructing pragmatic tree(s), and identifying themes and rhemes under the guide of thematization principle. This project aims to achieve automatical detection and modeling of informative intention at multi-levels and inference of collective intention, so as to accumulate a collection of intentions that will reveal the various connections among events. The corresponding research work will investigate novel research directions in reasoning and trend forecasting of events and enrich the text mining research at the discourse level.
英文关键词: Information extraction;Text mining;Knowledge discovery;Natural language understanding;Natural language processing