Miyabi

Each medium has its own information clues which can hardly be described by alphanumeric representations (we call them “media information clues”). They are, for example, shape, color, construction for still image, motion for movie, and tone, melody for auditory data. However, hypermedia system designers so far have been forced to assign many keywords or many links among objects in order to provide flexible and user-customized navigation. As a result, costs for organizing such systems rise accordingly, and it is difficult to apply this kind of alphanumeric approach directly to large-scale systems. The translation from media information clues to alphanumeric representation is clearly a barrier for the interaction between the system and its users. Instead, users should Ibe able to directly make a query to the hypermedia using the same representation as the object they want to retrieve. Media-based navigation is a newnavigation style for hypermedia systems. Users can use media information clues in both browsing and querying phase. Users retrieve objects using media information such as shape, color, motion, etc. This type of media-based retrieval becomes even more powerful as users can input the queries to the system as objects they perceive in the real world. The system, then, interprets the queries and shows possible candidates that match the query. Users browse these candidates, and show interesting objects to the system. The system finds all other objects that might be of interest to the user. Miyabi is a prototype of a navigation-based hypermedia database system, It is based on an extended EntityRelationship model and provides a wide variety of navigational tools such as schema browsers and media-based browsing tools. It is designed to clearly separate conceptual information from media processing so as to simplify its architecture. It is also designed to consider high level semantic structures to handle large amounts of hypermedia data. Miyabi has wide use of database conceptual modeling as well as user-friendly hypermedia operations. We have developed a Paris guide system and an electronic art museum to demonstrate the usefulness of Miyabi. Both prototypes have approximately 600 instances, 2000 links and 200 images based on six entities. The images in this system are, for example, paintings of art, landscape photographs of Paris, snapshot photographs of Paris, and so Permission to COPY without fee all Or part of this material ia that copying is by perrniseion of the Association for Computinu granted provided ttmt the copiee sre not mede or distributed for Mechinery. TO COPY otherwise, or to republish, requires e fee direct cornmercisi adventege, the ACM copyright notico ●nd th~o wtdlor epecific permission. title of the publication and its de(e ●ppear, ●nd notioo is @ven O 1993 ACM &89791-&~7/93/()()11, . *$1,So Hypertext ’93 Proceedings ;~33 November 1993 on. Users can navigate through the system based on many kinds of information clues including media information clues. In order to provide media-based browsing, we introduced the concept of the picture index. The picture index is an internal representation of the stored graphical object. The system generates a picture index automatically, by applying edge detection techniques and region divide techniques which are based on human observation of visual perception processes. At the current time, the picture index size is 24 x 24. It is composed of several regions, each of which has color and textual information. These characteristics are enough to represent the overview of the image. They are suitable to compare the user’s uncertain memory about the image, quickly. Users and system designers do not have to assign any keywords or index terms for content retrieval. Miyabi provides the following navigational interfaces: Hypertext navigation: This is the basic technique for browsing hypermedia systems. Users can extract many types of data by following links, for example, explanations, images and figures. Schema browsing: Users can browse through the hypermedia system based on the extended Entity-Relationship model. They can easily inspect the whole elements in the Entity, check the relation, and drift other entities. They can get the position of the data in the database by dragging it. Retrieval by alphanumeric query: Objects are retrieved from the alphanumeric queries using the database function. For example, users can retrieve images from the title, the author’s name, the museum and so on. Users may pick the item on the display and drag it into the area for the retrieval to input these alphanumeric clues. Retrieval results are displayed to the user simultaneously, Retrieval by visual query: Users can retrieve some images simply by drawing a rough sketch and showing it to the system. The system searches for the best match candidates by using picture index pattern matching algorithms. The system considers the characteristics such as shape, position, structure and…