Some Practical Observations on Integration of Web Information.

The patient showed an activated fibrinolytic system and was given dextran in order to prevent further platelet aggregation and fibrin deposition, and freshly frozen plasma and fresh blood were given to replace the coagulation factors.

Abstract

For past several months, I have continuously maintained two Web-based information integration systems. Collectively, these systems support database-like queries concerning the information found on several thousand Web pages from about 50 sites. My experience in building and maintaining these systems has led to certain practical observations. (a) Approximate extraction and approximate query-answering make information integration at this scale feasible; however, better tools are needed to make it possible for non-experts to create and maintain such a system. (b) More time and energy is spent in maintenance than in initial development. This suggests certain new research problems -- e.g., developing automatic means for maintaining wrappers, as opposed to creating them. (c) There is a tradeoff between the number of sources that are integrated, and the depth of meta-knowledge available about any individual source.