The popularity of several of the retrieval interfaces (10 most recent, calendar, and full text search) was also investigated. The web server access logs for April to June, 1998 were analyzed and the frequency of use of each of the three interfaces was recorded. This analysis showed that the ten most recently captured documents, the calendar, and full text search were the three most popular interfaces. These results suggest that users can readily adapt to new (ten most recent and calendar) retrieval interfaces. One potential reason for this is that the new interfaces presented here are intuitive, easy to use, and fast

4 Conclusions

A design was presented for a document storage and retrieval system known as the infinite memory multifunction machine (IM³). The concept for this approach includes capture of all printed, copied, or faxed documents and easy retrieval with a web browser. The capture of all documents processed by users, as a side-effect of their normal processing, eliminates the need for them to decide whether they should save the documents. This almost guarantees that a document will be present at some point in the future when a user needs it. This helps solve the problem of lost documents since when a user determines a document is lost, he or she can be assured the IM³ has a copy of it.

An implementation of the IM³ in a research lab with about 20 users was described. Over the course of more than two years, over 40,000 documents have been captured. A total of only 9 GB of disk space has been used. Today, this much disk space can be accommodated on a single 3.5 inch device costing just a few hundred dollars. The utility of this prototype implementation was determined by analyzing the web server access logs. It was shown that copied documents were accessed more frequently than printed documents. However, users still accessed old documents (more than two years old) that had originally been printed. This illustrates the utility of capturing printed documents without asking since it is unlikely that a user would have manually entered them in an image retrieval system when they were created.