D. Terence Langendoen, Brian Fitzsimons and Emily Kidder, University of Arizona

The GOLD Effort so Far

In this presentation, we describe how we have gotten to where we are now in the development of the General Ontology for Linguistic Description (GOLD) within the E-MELD project. The University of Arizona group, which was originally charged to work on markup recommendations for machine-readable materials in and about endangered languages, quickly realized that we stand a better chance of gaining community acceptance of standards if we do not attempt to impose detailed encoding rules for various kinds of linguistic materials, but instead create programs and tools that would enable machines to interpret what linguists mean by the encodings they provide. That is, we decided to follow the lead of the developers of the "Semantic Web", if not in practice, at least in spirit. We were also inspired by the design of OLAC, which supports resource discovery based on a shared semantics for document types, language names, and the like, designed by the entire research community.

Our challenge was and continues to be, how can we help linguists expose not only their metadata, but also their actual data and analyses on the Web in ways that make them comparable. Our first major effort was to build a structure -- GOLD -- for housing and relating fundamental concepts of linguistics about which there is broad consensus in the linguistic community. In this way, linguists who use those concepts in their analysis can relate instances of them in their documents to their occurrence in GOLD, thus preserving the integrity of the original analyses while making them all comparable.

We are now at the point where we need to turn the work over to the community at large. We propose to construct a framework in which linguists around the world can collaborate to extend the conceptual superstructure that GOLD represents, and to create services and tools that support the creation of much richer linguistic documentation on the Web than we can currently provide or even imagine.