Building wrapper agents for the deep Web

Abstract

The Web has rapidly expanded not only as the largest human knowledge repository, but also as a way to develop online applications that may be used for many different tasks. Wrapper agents may automate these tasks for the user. However, developing these wrapper agents for automating tasks on the Web is rather expensive, and they usually require a lot of maintenance effort since Web pages are not strongly structured. This paper introduces two programming languages for automating tasks on the legacy Web developing low cost, robust wrapper agents that may navigate the Web emulating browsers and may process data from the Web automating some user's behaviours. These languages are based on formal methods and W3C standards and are suitable for the legacy deep Web, automating tasks like data mining or information integration from different sources. A platform providing execution support to agents developed in these languages has been provided. © Springer-Verlag Berlin Heidelberg 2003.