Automatic filling in a form by an agent for web applications

The number of end-users using the Internet is increasing. End user initiated development of applications has become important for the automation of tasks. This paper describes a multi-agent framework for MOON (multiagent-oriented office network) systems, which are distributed and include window work in electronic commerce. Since window work is considered as a metaphor of the interface between Web service providers and requesters, this paper is focused on automatic filling in of forms by a user agent in collaboration with a broker agent, a key technology for electronic forms defined as HTML documents. A satisfactory success rate is achieved by using cognitive rules based on form layout. That is, the case rule of {IF #case THEN #action} is introduced. The #case is composed of the four attributes and their values based on cognitive information on the upper left, right and lower sides of the input field. The ontology of concept names is introduced for different expressions of the same meaning. Furthermore, by reasoning of similarity on incomplete matching of the case parts of rules, the success rate of automatic filling is greatly improved. The success rate is improved further to 100% by using experiential rules of other users, which the broker agent collects.