项目名称: 基于多源信息融合的元数据自动抽取方法研究
项目编号: No.61202232
项目类型: 青年科学基金项目
立项/批准年度: 2013
项目学科: 计算机科学学科
项目作者: 高良才
作者单位: 北京大学
项目金额: 21万元
中文摘要: 如何从非结构化或半结构化文本中自动获取元数据信息,即元数据抽取问题,是当前数字图书馆乃至整个信息服务领域的研究热点与难点之一。现有方法仅依赖文档本身的内容信息,难以逾越信息缺失与自身内容错误等障碍,不可避免地要引入大量人工审校,对抽取结果进行修正和补全。为此,本项目拟研究基于多源信息融合的元数据抽取方法,通过挖掘文档和外部数据的关系,构建多来源元数据信息的搜集与融合机制,充分发挥外部数据对抽取结果的修正与补偿作用,实现元数据的准确、全面抽取,突破现有方法的局限性。具体地,本项目将围绕种子元数据的生成、外部元数据的搜索、多源元数据的融合等关键问题,研究基于组合优化策略的种子元数据抽取方法、具有自适应性的元数据搜索策略、基于能量最小化模型的元数据信息融合算法、基于统计反馈的数据源质量评估体系等,为元数据抽取提供一个新的手段。其研究成果将大幅度提高元数据采集技术的自动化水平。
中文关键词: 元数据;信息抽取;信息融合;信息检索;
英文摘要: How to capture the metadata information from the unstructured and semi-structured texts, namely metadata extraction, is one of the major challenges and focuses in digital library, and even in the whole information service domain. The existing methods on metadata extraction primarily rely on the content analysis of texts. However, the results from such content-based methods often contain text errors and the extracted metadata is only a small part of the relevant metadata of resources. As a result, heavy manual correction and enrichment is needed to obtain accurate and complete metadata. Therefore, our project proposes an automatic metadata extraction method based on multi-source information fusion. Depending on the relationship between resources and external data, our proposed method first constructs the searching and fusion mechanism of multiple sources of metadata, and employs the metadata from external data sources to correct and complement the extracted results of content-based methods. Also, our method can break through the limitation of the existing methods, and achieve much more precise and comprehensive metadata. This project would focus on multiple key problems, such as seed metadata generation, external metadata search and collection, multi-source metadata fusion, etc. Overall, our project will have the
英文关键词: Metadata;Information Retrieval;Information Fusion;Information Searching;