?

Log in

No account? Create an account

Backwards | Onwards

About This Blog

When I first started working with ontologies, I think what I was really doing mentally was more along the lines of creating rather complex catalogs that a known set of instances would be structured by. And since my initial work in ontology was modeling the online publishing world I worked in and the future we hoped to create for it, we already had a lot of instances and shared taxonomies that we wanted to organize in the ontology: KB articles as part of the content model, Language and Geography coding that drew from ISO standards, the corporate product taxonomy, the roles of content producers, and company sites users came to.

The more I've learned about ontology, especially what Dean Allemang and James Hendler refer to as "ontology in the wild", I realize that the real strength of an ontology is its ability to provide a framework of meaning for instances that have yet to be identified or defined, or which are only partially defined, or defined in some private nomenclature. It acts as a kind of sorting tool for the raw noise of Web data, or any data source, and integrates it into its network of meaning based on how well a computer can use the relationships, restrictions, and taxonomies that make up the particular ontology.

This is why the ontologists I worked with would state, "Building an ontology is an iterative process." And it explains why they would get so frustrated at the slowness of tools that imported and reinferenced updates to the ontology. You have to do a lot of that when building an ontology and you don't want to have to wait for hours at a time. You have to poke around in an ontology, test it, try to break it, in order to ensure that it's flexible enough to meet the demands of a world that's always changing, yet accurate enough to turn the noise of the Web into valid meaning for humans.

This is a change in perspective for me. I understood it superficially in the workplace. It was the very flexibility, the ability to modify an ontology to meet changes in business strategy, that made it first so attractive to me as a solution to the problems we had in our content management space. But I assumed that individuals we organized in the ontology would have an appropriate set of correctly defined metadata so that they would all be correctly catalogued, and thus, findable when needed. Ontology as library system. It's not wrong. It's just not the whole picture.

Reblog this post [with Zemanta]