UI for a comparison of Submissions accross tasks / datasets

[mam10eks]

I saw a very nice way to visualize submissions across multiple tasks / datasets in the entity linking tool GERBIL.

The UI of Gerbil had the following main features for this:

• Select datasets (default: all)
• Select submissions (default: all)
• Select measures (forgot what the default was, maybe all?)

For this, Gerbil rendered a table with the main results but also some figures (e.g., task/dataset on the x-axis sorted by decreasing effectiveness and effectiveness on the y-axis resulting in a line plot.)

[mam10eks]

This would be especially interesting for TIRA/TIREx (cc. @potthast, @heinrichreimer, @seanmacavaney).

@juhehehe, @Kavlahkaff, If you want, you can already start a bit with brainstorming on the UI for this.

[potthast]

Can you post some screenshots? It's not clear to me where to find them on the linked page.

[mam10eks]

I also could not find the screenshots anymore.

Here is a sketch with what I remember from the talk:

Does this sketch clarify the idea?

I would really love to have something like this.
Maybe this would be a good next todo for @juhehehe?

[mam10eks]

So in the end it is basically a table.
There was also a way to draw the content of the table in a figure, but the table would be the main thing to me.

[janheinrichmerker]

Yes, I agree, comparing the results of a bunch of selected approaches in this table-like fashion would also be the main use case for me.

Further nice-to-have's:

• CSV export
• online significance tests (for starters just a t-test with fixed alpha, that automatically applies multiple test correction based on the number of selected approaches)

[mam10eks]

I finished the REST endpoint for this and integrated it into the existing RunList.vue:

From that, I think @juhehehe can finalize the UI of this.
(You see this table on the submission pages)

[github-actions]

This issue has been marked stale because it has been open 60 days with no activity.