Integration Between Hypermedia Information Retrieval System and Information Retrieval Systems

In hypermedia systems, links are used to connect related information in different types of media, for example: text, images, sound, etc. In this way, authors and users have means of accessing the information in a non-sequential manner, in the way they prefer, by following links that may be visualised as highlighted text or by other means. However, when simply following links while seeking information, users may become disoriented and frustrated at their lack of ability to find specific or useful pieces of information. This is why the integration of information-retrieval techniques into such systems makes sense. The search for information can then be as well satisfied by specifying an information-retrieval query. Some systems already show this integration between hypermedia and informationretrieval systems. However they are usually only capable of text retrieval (Frei et al, 1995) or, when they have real content based multimedia information-retrieval capabilities (Lewis et al., 1996), the efficiency and effectiveness are very low. To overcome this problem we have developed a hypermedia information-retrieval integration solution that uses automatically created text descriptors to achieve multimedia information-retrieval using text retrieval techniques (Dunlop et al., 1993). In this work, there were two main aspects under consideration: The automatic creation of text descriptors for multimedia documents, using information taken from the hypermedia network both at the document level and the abstract classification level. The type of information considered as part of this process includes link types, link descriptions, link anchors, keywords and logical classification concepts. -The integration of open hypermedia systems, in this case the Microcosm system (Davis et al., 1992), with text information-retrieval tools to allow multimedia retrieval using text descriptors for non-text media. After the implementation of this solution we presented some results and conclusions about the effectiveness of such a system. Supported by the conclusions and after further suggestions found on research where navigation through concepts was possible (Crouch, 1989, Cunliffe, 1997) we then present the skeleton of a model. For this model major concern is placed on the integration between structure information from hypermedia and content information from information retrieval systems, both at the document level or at the abstract level of navigation. This will allow the creation of a new system where there is no clear distinction between navigation and querying, which ever the media is considered. Implementation deadlines and testing suggestions are given at the end before overall we point out some conclusions about the research that we develop. 1 This research was supported by PRAXIS XXI PORTUGAL under grant BD/5399/95