Until now, I have had a simple system that uses several agents to parse a WSDL and drop its contents into a database, with some other agents that analyze the individual input and output parameters for each service and try to find matches for them among the parameters already in the database. Each match is loosely ranked based on how exact a match it is to any other parameter.

As of this evening, I am finally doing something with those scores. I added code that will take the set of matching parameters and their scores, apply a configurable weight to each type of match, and evaluate them to find the highest scoring match. That match is then stored in the database with the original parameter. This should give me some reasonable progress to show my advisor the next time we meet. It’s not much, but it is enough of a framework that I should be able to add an arbitrary number of evaluator agents and have all their scores accounted for and evaluated.

Now comes the tricky part: figuring out whether all these matches form any sort of reasonable ontology.