Andrea Preite Martinez wrote:
>
> If you don't mind, I'll post your file on the twiki page of the IVOA UCD
> working group (http://www.ivoa.net/twiki/bin/view/IVOA/IvoaUCD )
> as an example of application of UCD-tools.
>
> I was thinking of organizing a session at next IVOA InterOp meeting in May
> on applications of ucd-tools to real cases and feedback from users.
> I hope you'll be there to present your work.
>
We're certainly plan to do so. Feel free to post our paper where you feel it's appropriate. Note that the testing of the tools was really the secondary issue for us. Our primary concern was whether the UCDs could adequately describe the content a table column and only then how we found that appropriate UCD.
> Now, in about 100s I can assign UCDs to more than 40.000 columns.
> But the real problem is not only time (because you have in any case the
> control-time to consider!).
We'll certainly be trying out the tools you describe. Even for our measly 400 tables it's quite a slog to get get UCDs for all tables. One thing that we will be looking at is how well the UCDs we generate compare with yours on a few of the tables we hold in common. Generally I think we'll defer to your selections there, but it will be an interesting experiment to see how well we match for at least a few test cases. (I think there is one common table in our trial set but I haven't looked at this yet.)
Not sure what you mean by control time -- is that the time you take to check the results? In that case I think I agree that that's the real rate limiting step. Today I think that a human has to look at every UCD to confirm that it is appropriate.
Tom Received on 2006-01-27Z14:37:28