Easy-to-use methods for retrieval of stored documents are also a key factor for success in user acceptance of such a system. This is especially the case since by reducing the filing-time effort almost to zero, users may need to perform additional tasks at retrieval time. Our objective is to reduce this overhead as much as possible.

Because of the relatively large volume of documents that are captured, it was deemed necessary to provide multiple, complementary retrieval techniques. Following the themes of ease-of-use and familiarity, retrieval interfaces were created that utilize paradigms users encounter daily. For example, one method displays thumbnails of document images in a calendar along with appointments. Users can retrieve documents by browsing the calendar for images that were saved at some time near that of an appointment. Other retrieval techniques that have been the subject of our research include image-based methods for document retrieval [1]. The detection of duplicate documents is another obvious concern, both for storage reduction and version detection in a user interface. Text-based and image-based techniques for duplicate detection have been developed [4, 5, 7].

Several document image storage and retrieval systems relevant to the IM³ approach are described in the literature. An approach for maintaining a personal database of scanned scientific papers has been described [8]. An approach designed for storing and retrieving scientific papers in a client-server environment was discussed in [9] and the need for sophisticated retrieval techniques was presented in [2].

The rest of this paper describes a prototype implementation of the IM³ system concept. The copiers and printing system in an office with about 20 users were modified so that every document they processed could be automatically saved. Captured documents were stored on a central server and made available for retrieval with a web browser. Provisions were made for selective encryption and security of users' databases in a way that promoted sharing information among users. A variety of retrieval and document communication interfaces were provided. User interactions with the retrieval interfaces were recorded and analyzed to determine the popularity of the different interfaces.