Designing User-Interface for a Digital Library with Concept Learner Assisting Users to Find the Required Documents

当文档/查询项在概念层次中被它们的意义替换时，可以解决信息检索中的词汇问题，并且可以很好地捕获文档/查询内容。关键是如何指定适当的意义。首先提出了一种处理术语义歧义的方法，该方法通过义项之间的m-to-L对应关系来捕捉查询和文档之间的概念关系，从而只选取有助于保持较高检索性能的义项来简洁地表达文档内容，并对其中的概念层次进行编辑，以反映所有用户共同的概念关系，从而不反映个人的思维方式。其次，我们提出了一种信息检索范式，将待检索的文档以数字图书馆的形式收集起来，并帮助用户在数字图书馆中进行浏览，以检索满足其需求的文档。通过基于实例的学习算法获取辅助信息，交互地收集相关/无关文档，并在计算机/信息科学和医学科学两个测试集合上进行计算和统计实验，充分验证了所提方法的优越性。还设计和建造了一个数字图书馆及其导航器，以最大限度地减少用户浏览的感知负担。

The vocabulary problem in information retrieval can be dissolved and the document/query contents can be well captured when the document/query terms are replaced by their senses in a concept hierarchy. The point is how to specify the appropriate senses. We first formulate methods of treating term-sense ambiguities, which select only the senses that contribute to keep the retrieval performance high and express the documents contents concisely, by capturing conceptual relations between a query and a document through m-to-l correspondences of the senses.The concept hierarchies in the above are edited so as to reflect the conceptual relations common to all the users, and therefore they will not reflect the individual way of thinking. We secondly proposed an information retrieval paradigm, where the documents to be examined are collected in a form of a digital library and a user is assisted when he/she browses in a digital library to retrieve the documents meeting his/her requirements. The assistance information is caught by an instance-based learning algorithm, which collects interactively the relevant/non-relevant documents.The advantages of the proposed methods are fully tested by computational and statistical experiments on two test-collections as to computer/information science and medical science. A digital library and its navigator are also designed and constructed so that the user's perceptual burden for browsing is minimized.