[16:15:37] <miriam_>	 isaacj :) happy Friday!! I have  a question - can we use the Wikidata topic model to generate similarity scores between Wikidata items? Are the underlying embeddings avaliable somewhere? thanks :))
[16:16:53] <isaacj>	 miriam_: good question -- yes, we could but the underlying embeddings aren't stored anywhere (they are generated as the hidden layer in the fastText model but can be extracted)
[16:17:25] <isaacj>	 how many items do you need?
[16:18:49] <miriam_>	 isaacj, so the use case is the following. For the project on understanding the role of images in Wikipedia, we are asking reading comprehension questions on Wikipedia articles. We can automatically generate questions with the correspoinding answers, but we need to generate alternative (wrong) answers to populate a multiple-choice questionnaire
[16:19:54] <miriam_>	 so I thought, for those answers which are actual Wikidata items, e.g. "Lorenzo de Medici", we could automatically look for the most similar entities in Wikidata and see if this gives good alternatives
[16:20:20] <miriam_>	 so to answer your questions, potentially all of wikidata :O
[16:20:38] <isaacj>	 ok, that's fine (in some ways easier though obviously takes a while to process)
[16:20:44] <isaacj>	 any languages to filter to?
[16:21:42] <miriam_>	 oh, right, we have English only for now (sadly)
[16:22:31] <miriam_>	 no rush isaacj, we still have to run the first pilot with manually generated questions :)
[16:23:21] <isaacj>	 kk -- i have a script that is meant for bulk predictions so i'll just modify that for english and outputting the embeddings instead of predictions. shouldn't be too hard so can get it to you in the next week or so as a SWAP notebook and then if you need to tweak, you'll be able to do that easily
[16:54:17] <miriam_>	 thanks isaacj that would be super helpful, but really, noO rush!!
[16:57:56] <isaacj>	 :thumbs up:
[17:52:00] <leila>	 o/
[17:52:49] <leila>	 fkaelin: shall we connect some time today for finalizing some of the steps we've been talking about? I should have my part done by noon PST.
[18:06:41] <fkaelin>	 leila: sounds good, i am looking  it over now. let me know when it's good for you to meet
[18:08:37] <leila>	 great. inviation on your way.