Return top-k candidates for ED?

[AVBelyy]

Hi,

Is it possible to return top-k most likely candidates for the disambiguation step? This would help in my application: when the top-1 predicted entity is incorrect, we give users a chance to correct that.

Thanks!

[mickvanhulst]

Hi,

At the moment this is not implemented, but it should not be too much work. If you use our API, you can see in our server file (code) that we are using a function called process_results (code). This function receives our predictions. Our predictions dictionary should include both candidates and scores as a key respectively. Although we currently don't return them, it should only take a few lines of changes in the process_results function to get the top-k predictions that you want.

If you have the time to work on this, it would be great if you could file a PR as I can imagine that this is a useful feature for many applications :)!

[aflah02]

@mickvanhulst Is this still open? I have some free time on hand to take a look!

[KDercksen]

@aflah02 it is still TBD! It would be great if you can take a look :) See also here for the dictionary keys.