Datasets: agree on useful tags

[ajdapretnar]

Orange version

3.11.dev

Expected behavior

Data sets are marked with useful and standard tags.

Actual behavior

Data sets are marked with random tags or useless tags.

Steps to reproduce the behavior

Open Datasets. Wine data set has tag 'wine', zoo data set has tag 'animals'. Not giving any extra information.

Additional info (worksheets, data, screenshots, ...)

I propose a solution for improving tagging:

• tag for purpose [classification, regression, clustering]
• tag for special attributes [geolocation, time, image]
• tag for add-on it should be used with [bioinformatics, timeseries, image analytics]
• field it covers [economy, social sciences, genomics]
• special format [basket, multitarget]

Please add your suggestions.

[ajdapretnar]

I would like to implement this. @BlazZupan @lanzagar What do you think?

[lanzagar]

I agree tags could and should be improved. The proposed tags look good enough (better than current). I am not sure about the usefulness of the first item - tags for purpose - isn't that information visible in the Target column (categorical=classification, numeric=regression, no target = unsupervised/clustering).

Otherwise I think you can start a PR with tag improvements.

[ajdapretnar]

I am not sure about the usefulness of the first item

I agree. I have prepared a PR. I am not sure the solution is optimal, but at least now it is fairly consistent.