Building: Grand Hotel Mediterraneo
Room: Africa (formerly America del Sud)
Date: 2013-10-30 03:00 PM – 03:15 PM
Last modified: 2013-10-08
Abstract
Metadata disorder, maintenance costs, and cross-walking complexity are increasing due to the number of scientific data schemes and standards (Dublin Core, Darwin Core, OWL Web Ontology Language, etc). SeaIce*, a multi-domain community-driven online metadata dictionary, is introduced as a potential solution. Users would be researchers trying to describe their datasets, who would choose SeaIce as an alternative to picking among dozens of incomplete or outdated standards. Without logging in, they would search for and insert terms into metadata that they’re creating or modifying. If unsatisfied with the terms they found -- or didn’t find -- they can take action, by registering and logging in, to create new terms, edit their own terms, and comment on, and vote up or down the terms of others. All domains and all parts of metadata speech reside in SeaIce, similar to a natural language dictionary. This includes element names, element values, units, and inter-term relationships (broaderThan, narrowerThan, “instanceOf”, etc). We expect existing vocabularies to be added in order to permit building on prior work. Because multiple competing definitions are expected to be common and to co-exist indefinitely, unique persistent concept identifiers are critical, and they can easily be repurposed for linked data applications. All new terms enter a Vernacular portion of the dictionary, and when stable, high-quality terms emerge in the meritocracy we have set up (reputation-based voting similar to StackOverflow), they move into the Canonical portion, which means that their evolution stops. End users, depending on their risk tolerance, will use a balance of Canonical and Vernacular terms. In a departure from common practice, we do not discuss serializations (e.g., in RDF or XML) because, while gratifyingly easy and concrete, they generate noise that hinders subsequent discussion.
* http://seaice.herokuapp.com/. The project name, SeaIce, was not meant to connote a domain of application, but was picked randomly from a data set we were studying at the time. We are looking for a better name.