支持分层细粒度分面检索的分面分类树构建方法研究

Almost faceted literature search systems take the subject category as a thematic facet, which cannot express the topic of literature specifically and precisely.It results in the bad fact that the oriented-topic faceted search degenerates into traditional search based on keywords, which brought about information-overloading problems. Consequently,the method of faceted taxonomy construction for hierarchical and finegrained literature search will be well investigated in this project. Contents of the project are as follows: 1. We will do the research of the oriented-topic faceted model for hierarchical and finegrained faceted literature search.2. We will solve the problems of the acquisition of facet terms and relations,and the optimization of faceted taxonomy according to the characters of oriented-topic literature faceted search. 3. The faceted taxonomy construction approahes are all applied to a prototype system. By means of simulations,the theory and approaches above are verified. The traits of this project are that a novel hierarchical and finegrained literature faceted search mode is proposed and the method of the oriented-topic faceted taxonomy is explored. The implementation of this project will enrich and improve the theories and approaches of information retrieval..Expectant fulfilment: publish about 10 high level papers; cultivate 2 Doctor Degree Candidates and 3 Master Degree Candidates; write 1 state invention patents; develop a prototype system.

现有文献分面检索系统大多以学科分类作为主题分面，不能精准反映文献主题内容，导致针对文献主题的查询，退化成传统的基于关键字查询，降低了文献主题检索的可用性，带来了信息"过载问题"。为此，本课题拟以特定领域的学术文献为研究对象，研究支持分层细粒度分面检索的分面分类树构建方法。内容包括：1.分层细粒度文献主题分面模型研究；2.结合文献主题分面检索特点，重点解决隐式分面术语及关系获取、分面分类树优化等问题；3.以Yotta系统中的文献资源检索为应用载体，研制原型系统，对所提理论与方法进行测试与验证。项目特色在于：提出一种分层细粒度分面检索的新型检索模式，探索面向文献主题的分面分类树构建方法。本项目有助于丰富与完善信息检索领域的相关理论与方法，实现基于主题关联的文献组织与导航，从而有助于缓解"信息过载"问题。.预期成果：发表学术论文10篇；培养博士生2名，硕士生3名；国家发明专利1项；研制原型系统。

现有文献分面检索系统大多以学科分类作为主题分面，不能精准反映文献主题内容，导致针对文献主题的查询，退化成传统的基于关键字查询，降低了文献主题检索的可用性。为此，项目组研究支持分层细粒度分面检索的分面分类树构建方法中的关键技术问题。内容包括：1. 层次化细粒度分面模型形式化表示，设计了支持丰富语义的交互式迭代导航操作；2. 针对Wikipedia条目页面特点，提出了一种逐层特征投票模型来自动识别领域术语；3. 提出了基于Motif的上下位关系的抽取方法；4. 基于拓扑特征的分面分类树生成；5. 为了实现基于分面分类树的分面检索，对文献自动链接算法进行研究，提出了基于数据重构的多层次标签传播算法以及基于ELM的多标签分类算法。. 支持检索结果分面分类的检索方法称为分面检索。分面分类树是分面检索的核心，可以用于资源的检索，尤其是非结构化资源的管理与访问。本项目的研究有助于实现基于主题关联的文献组织与导航，从而有助于缓解“信息过载”问题。