Dynamic Dictionary = Cooperative Information Sources

In this article a multi-source system is presented as a case-study in order to show how agents can be used to build a dynamic dictionary. In this context a dictionary has a much wider sense than usual: anything that can give information about a term (whatever term may be) is a dictionary. The dictionary is composed by several information sources with different domains (knowledge area and cardinality), relevance/richness and cost (cost may be a function of time, resources, communication traac, money, etc). An information source is an agent that, given a term, can return some knowledge about it, in an universal representation schema (feature structure). To answer the user's query about a term, the system aims at nding an information of maximum relevance at a minimum cost; this requires a working strategy (a compromise solution). Each information source agent provides meta information about its behavior so that a special agent |strategy builder| is able to deene a strategy. Working in this case-study we learned that sometimes, agents can be as simple as operating system commands (UNIX like) and aplications, properly encapsulated with an interface layer.