Research on the Kansei Retrieval Methods for heterogeneous media Information

本研究实现了基于感性的异质媒体信息检索方法。在该方法中，检索关键字和检索目标是广域网上可以访问的任意类型的媒体数据。在本研究中，对于每一媒体元数据，通过使用代表感性信息的词汇的自动提取方法，以词作为公共媒体来表示媒体元数据。本研究通过将这些信息映射到可以用统一的算子来衡量每个词之间的距离的通用空间中，实现了异质媒体的连接，体现了人的感性思维。本研究还通过在元层次上连接单个检索方法的元数据和媒体数据，实现了不同检索方法的互操作。该方法实现了对各种媒体具有多种检索功能的多媒体数据库，实现了在心理学研究的基础上，将每个媒体元数据映射到与人的感性思维相对应的宇宙空间。在传统方法中，重点是实现一种针对单一媒体元数据的高效检索方法，而该算子可以消除现有媒体信息检索系统中以词表示的感性信息之间的差异，该方法的实现使得通过连接现有媒体元数据来生成新的信息成为可能。此外，该方法扩展了搜索者对检索关键字的表达范围，使基于人的感性的各种信息检索如愿以偿。这种方法可以指数级地增加对现有媒体数据的访问和现有数据库的实用价值。

This research realized the heterogeneous media information retrieval method based on Kansei. In this method, retrieval keywords and targets are arbitrary kind of mediadata which are accessible from wide-area network. In this research, mediadata were represented with the word as common media by using automatic extraction methods of words representing Kansei information for each mediadata. This research realized heterogeneous media connection by mapping these information into the universal space which can measure the distance between each word with unified operator reflecting human's Kansei. This research also realized the interoperability of different retrieval methods by connecting metadata for individual retrieval method and mediadata on a meta-level. This method enables the realization of multimedia database which have multiple retrieval function for various kind of mediadata.In this research, the operator which can map each mediadata into universal space corresponding to human's Kansei based on psychological research was realized. In conventional method, the focus was realizing a retrieval method with high effectiveness for single mediadata, however, this operator can erase the differences between Kansei information represented by words in existing media information retrieval systems.The realization of this method enables the newly information generation by connecting existing mediadata. In addition, this method extends the searcher's range of expression of retrieval keywords and enables the various information retrieval based on human's Kansei just as the searcher intended. This method increases access to existing mediadata and utility value of existing databases exponentially.