Less-conscious Information Retrieval Techniques using User's Implicit Intentions

我们在意识和无意识之间的低意识层面开发了如下的信息检索技术。(a)基于用户交互提取隐含意图（a-1）利用媒体操作提取用户意图：我们开发了一种利用用户意图的检索方法，用户意图是通过不自觉的媒体交互提取的。（a-2）从用户的浏览行为建模和估计浏览模式：我们根据用户的目的强度定义了探索模式和漫步模式两种浏览模式。然后从浏览页面的内容和URL域两个方面分别定义了两个集中度。我们开发了一种利用集中度来估计用户浏览模式的方法。(b)基于自动查询生成的低意识信息检索方法（b-1）检索跨媒体内容的查询生成：我们开发了元模型，将（A -1）提取的用户意图传播到其他媒体内容的查看、收听、编辑和检索等操作中。然后，我们开发了原型系统并对我们的方法进行了评估。（b-2）基于用户浏览模式的用户辅助方法：我们开发了使用社交标签的推荐技术，用于在（a-2）中定义的漫步模式中帮助用户。为了找到要推荐的页面，我们首先从附加到浏览页面的标签中找到标签的组合，这些标签是父子关系。然后我们替换这些标签组合中的一部分，利用替换后的标签组合找到与浏览页面相似、一般化或相关的页面。

We developed information retrieval techniques on less-conscious level that is between conscious and unconscious as follows.(a) Extracting implicit intentions based on user's interactions(a-1) Extracting user's intentions using media operations: we developed a retrieval method using user's intention that is extracted by media interactions which does not conscious to information retrieval.(a-2) Modeling and estimating browsing mode from user's browsing behavior: we defined two browsing mode, an exploring mode and a strolling mode in consideration of intensity of user's purpose. Then, we respectively defined two concentration degrees form the aspects of contents and URL domains of browsed pages. We developed a method to estimate user's browsing mode by using the concentration degrees.(b) A less-conscious information retrieval method using automatic query generation(b-1) Query generation for retrieving cross media contents: we developed meta-model for spreading user's intentions that is extracted by (a-1) to other media contents' operations as viewing, listening, editing and retrieving. Then, we developed the prototype system and evaluated our method.(b-2) User assistance method based on user's browsing mode : We developed recommendation technique using social tags for assisting users in a stroll mode defined in (a-2). To find pages to be recommended, we first find combinations of tags, which are in parent-child relationships, from the tags attached to a browsing page. Then we replace a part of these combinations of tags and we find the pages which are similar, generalized or related with the browsing page by using the replaced combination of tags.