Research

This section is where the atlas’s own studies will be published. It has none yet, and it says so rather than filling the space: everything else here reports something that was measured — 299 audited cases in the Casebook, 16 model configurations over 3,266 questions in OenoBench — and a research section that opened with forecasts would be the one page asking to be taken on trust.

What is being worked on follows from what the two datasets already show. The Casebook records what has been deployed and how well the claim behind each deployment is sourced; the interesting questions start where that record thins out. Which deployments survive contact with a second season. What the gap between a pilot and commercial operation actually costs, in the cases where both ends are documented. Whether a benchmark score on wine knowledge predicts anything about a model’s usefulness in a cellar. Each of those needs data that one person cannot collect alone.

If you hold that data, please get in touch. Estates, cooperatives, laboratories, vendors and research groups all have the operational record these questions need, and a study built on it is more useful to the people who supplied it than to anyone else. Collaboration on a study, a dataset, a review of a draft, or a correction to something already published: [email protected].

Until there is something to publish, the measured work is in the Casebook and on the OenoBench leaderboard.