Early agent research recognised that co-operating
agents require access to unambiguous, semantic de-
scription of the same concept, entity or object. In
fact, agent-based research on this problem antic-
ipates many of the current initiatives of the Se-
mantic Web project. The proposed solution in-
volves developing a domain-specific ontology that
can be mapped to other ontologies as required.
In this paper we describe an alternative approach
which allows autonomous agents to index shared
objects without requiring ex-ante agreement on an
ontology. Using a process of distributed negotia-
tion, each agent builds a lexicon of the problem-
solving competences of other agents. We present
an overview of our work using this approach in
three domains: a web services scenario, a multi-
case-based agent approach and finally, Tagsocratic,
a blog-indexing service. We then describe our fu-
ture work on several open issues related to this re-
search.