[00:11:07] dsaez: I hope you're asleep but if you're not: I'm happy to be the dry-run attendee for your presentation. You don't have to tell me ahead of time, whenever you're ready and if you want to have input, just let me know 2 hours before that. ;) [00:11:35] and in return I will drag you for a dry-run on Monday or Tuesday for mine. :D [16:54:38] bmansurov: sooo, I'm reading the thread about the removal of hand-coded interlanguage links. [16:55:08] bmansurov: and my question is: do I understand correctly that we can't address the en:Neoplasm case under the new scenario? [16:56:06] bmansurov: basically, we won't have a way to know that en:Neoplasm is the same thing as de:Tumor, because there is not link to tell us en:Neoplasm is the same thing as Neoplasma which gets redirected to de:Tumor. [16:56:58] bmansurov: by "we don't have a way", I mean "we don't have the old way". of course, we can find some other ways to learn that Neoplasm in English is the same thing as Neoplasma in German. Here comes dsaez' magic. ;) [17:20:44] leila: yes that's correct. Currently, if there's no interlanguage link between articles, then the way described in the paper won't work. [17:21:32] leila: there maybe another way, I'm playing around with different techniques. Nothing substantial yet. [17:22:30] * leila thinks [17:27:01] bmansurov: do you get a sense that there is agreement that en:Neoplasm QID should be merged with de:Tumor QID and it's just the matter of someone doing it? if that's the case, that's an easier problem. [17:28:43] leila: do you mean merged on our end, or on Wikidata? [17:29:18] bmansurov: On Wikidata. Amir was offering in the phabricator task that a few of them help with merging if I understood correctly, right? [17:30:23] leila: I'm not exactly sure, because we already know of the connection between Tumor and Neoplasm, why hasn't anyone merged them? [17:30:40] leila: maybe he meant adding interwiki links, which are being phased out imo [17:31:00] bmansurov: basically we have two challenges. 1) the type of en:Neoplasm de:Tumor challenge that in the past we used to be able to address with manual ILLs and we currently can't. 2) the type of challenge we have with en:Cherry which seem to not have a de equivalent, which is of course incorrect as Cherry in German is covered under Vogel-Kirsche article, I believe. [17:31:54] bmansurov: yes. that's a good question and goes back to whether there are reasons to keep them as two separate entities (are they actually different from the scientific perspective?) [17:32:30] btw, point (2) that I mentioned above is harder, please just keep it in mind, and let's come back to it with dsaez. [17:32:39] leila: I see [17:33:19] OK, makes sense [17:34:03] I don't know how to handle point (2) without a pipeline to feed back to the model: we can do some synonym detection on our end, but at the end, we will miss things, and there should be a way for the user of the API to tell us "this thing actually exists" and we remove it from recommendations. [17:34:57] bmansurov: I think we should set up a call with dsaez and discuss this in more depth. can you do that? [17:35:23] leila: sure, I'll send an invite. We should probably talk about resolving (1) first, though. [17:36:21] bmansurov: re (1): can you check with Amir if this is just the matter of merging QIDs (in this specific example case) or there are reasons these two should be kept separate? if there are reasons, then we may be able to model those as well. [17:37:50] bmansurov: if we know this is just the matter of someone merging it, we can work with Magnus to surface QID merges through one of his tools [17:38:56] bmansurov: the "just" of course refers to finding a clever way to identify them first. ;) [17:38:59] leila: invitation sent [17:39:37] bmansurov: yup. thanks. [17:39:42] leila: OK, I'll see what Amir says [17:40:03] bmansurov: thanks. this part is a bit hairy, but it's important. let's get to the bottom of it. [17:43:20] I'm lost [17:47:06] dsaez: we may need your help in improving the model behind the article recommendation for creation API. [17:49:30] dsaez: there are two problems: 1) There are QIDs that are about the same concepts and need to be prioritized for being merged. 2) There are QIDs that are not the same, but the content in their corresponding Wikipedia articles can make them very similar. For example, en:cherry seem to suggest that de doesn't have an article about cherry (no interlanguage link to de), while the concept of cherry is covered, I believe, in de under de:Vogel- [17:59:00] abartov: do you have 20-min brainstorming time for an upcoming presentation I have? No worries if you don't. :) [18:01:47] bmansurov: ow boy: T203041#4646249 ;) [18:01:48] T203041: Output 2.1: An improved task recommendation API - https://phabricator.wikimedia.org/T203041 [18:01:59] leila: yeah reading that [18:09:13] leila: sure [18:10:47] abartov: thank you.