Wikidata:Requests for permissions/Bot/ImplicatorBot

The following discussion is closed. Please do not modify it. Subsequent comments should be made in a new section. A summary of the conclusions reached follows.

Regretfully closing as withdrawn by operator. Magnus has published the code under the GPL license, so hopefully someone else is willing to pick up this task. Legoktm (talk) 02:43, 10 April 2013 (UTC)

---

ImplicatorBot

Operator: User:Magnus Manske

Task: Bot account used for adding implicit statements.

The bot can automatically add implicit statements. If item A says "father:B", that implies that A and B are people, B is male, and has a child A. The bot will not add claims for an existing property at the moment, so it can't add e.g. multiple gender. I made some test edits with the bot. --Magnus Manske (talk) 20:49, 19 February 2013 (UTC)

•  Support I do not see any possible issues with the task.--Ymblanter (talk) 21:03, 19 February 2013 (UTC)
•  Oppose The task is pretty logic and deterministic and will avoid a lot of manual edits. However you got to find a way to undo your edits if the property you started from is removed --Thieol (talk) 21:03, 19 February 2013 (UTC)
  •  Comment Not sure I understand. What would be the difference to "reverting" a manual edit? If A is the child of B, and B is the father of A, and you want to remove both connections, then that's what you have to do. There should be JavaScript to automate this; however, I don't see what it has to do with this bot. --Magnus Manske (talk) 00:01, 20 February 2013 (UTC)
    •  Comment Magnus, I think that the use of logical assertions is a really very good idea and my opposition is due to vandalism,errors and immature properties. wikidata is not mature enough for the moment and your bot will be very useful as soon as wikidata information and specification become more reliable--Thieol (talk) 00:01, 20 February 2013 (UTC)

•  Comment I think will be good idea to check related article for usage of person-related infoboxes. Property could be added by mistake or as vandalism. --EugeneZelenko ([[User
  • Of course, because there never are mistakes or vandalism on Wikipedia! :-) Parsing infoboxes is not trivial, and often such information is missing. Also, shouldn't we be standing on our own feet here? --Magnus Manske (talk) 08:57, 20 February 2013 (UTC)
•  Question Could you perhaps program the bot to re-check its work occasionally, and see if any claims it based its work on have been removed? (And then log them somewhere.) To me, this seems the best way to avoid replicating vandalism. (Obviously the bot is going to wind up replicating some vandalism, and if there's no way to avoid it, then I'll support anyways, since I think the pros outweigh the cons, but it'd be nice if you could put something like this in place.) — PinkAmpers&^{(Je vous invite à me parler)} 20:10, 24 February 2013 (UTC)
  • Or introduce a time delay for the vandalism to be reverted. --Rschen7754 09:19, 26 February 2013 (UTC)
•  Support I think it works Mojdeh.h (talk) 10:05, 25 February 2013 (UTC)
•  SupportMonobit (talk) 17:21, 26 February 2013 (UTC)
• Support this was something I was hoping would be easily implemented. And it looks like it has been. Excellent. The Rambling Man (talk) 21:48, 27 February 2013 (UTC)
•  Neutral maybe very useful but perhaps very dangerous. It could be better if the bot, instead of editing items, would simply create logs somewhere and then some users could check and update them (maybe with a gadget) --Ricordisamoa 21:52, 27 February 2013 (UTC)
  • I don't think that is feasible. There are going to be thousands of items that will need this, which humans simply won't be able to keep up with. Legoktm (talk) 02:46, 28 February 2013 (UTC)
•  Comment This request has received no comments in 5 days. I will close it as approved in 24 hours if there is no opposition by then. Legoktm (talk) 05:31, 4 March 2013 (UTC)
  • Could we at least give the operator a chance to respond to the concerns regarding vandalism? That's the one concern that I have. --Rschen7754 05:51, 4 March 2013 (UTC)
    •  Comment Legoktm, im my opinion, you should not approve it so quickly. Too many neutrals and comments, even if i know how frustrating it is to wait for bot approval. Thieol talk 05:51, 4 March 2013 (UTC)

 request please do 200 edits to see how much is it precise?▬ Reza1615 / T 09:29, 5 March 2013 (UTC)

• What I would like to see is an exact list of which statements will be implicit. It's fine (with me) if the list is easily changed to reflect the code/future requests, but either way, that should be documented for normal editors to see. --Izno (talk) 18:53, 6 March 2013 (UTC)
•  Support, but please make a list of all implications you're using and let them be checked before using. For example your example can go wrong, there might be also famous animals in Wikidata, so father does not implicit that both are humans. -- MichaelSchoenitzer (talk) 23:39, 7 March 2013 (UTC)
•  Support I think it's not the task of this bot to test, if properties of items correct are. So in my opinion, the Bot can get the botflag.--CENNOXX (talk) 13:10, 20 March 2013 (UTC)
•  Comment Are you (the operator) still interested in pursuing this task? The trials have not yet been done, and there have been no comments from you for over a month now. — Hazard-SJ  ✈  02:18, 24 March 2013 (UTC)
•  Oppose for now. First of all, I hope no bot would be approved based on only 5 test edits. I also concur with MichaelSchoenitzer that the example given seems to be overly naive. I think a bot should not make edits based on implications unless the community has a chance to discuss them first, to improve the chance of identifying such errors. My other concern is that the edit summaries do not allow for easy manual checking. One has to guess what logical assumptions the edit is based on. I don't think the bot should be approved unless the edit summaries are improved, or the bot records its edit rationales in a log page on-wiki (and preferably both). --Avenue (talk) 09:40, 24 March 2013 (UTC)
•  Support Good idea. BTW Happy Magnus Manske Day|. --  Docu  at 11:44, 24 March 2013 (UTC)
•  Comment I shall commence test edits within the next few days. --Magnus Manske (talk) 10:41, 25 March 2013 (UTC)
  • Thanks! — Hazard-SJ  ✈  02:55, 26 March 2013 (UTC)
    • Did some test edits. --Magnus Manske (talk) 20:28, 27 March 2013 (UTC)
•  Comment it will help a lot by saving time. I suggest, when property 'A' is connected to property 'B', each time discussion should be carried out to avoid any possible loopholes in connecting them. In some case it may need some conditions like , if property 'C' is there than 'A' should be connected to 'B'. In general, I support this Bot but with conditions.--Nizil Shah (talk) 21:19, 27 March 2013 (UTC)
•  Oppose, as there seems to be little to no documentation on how this bot would work. Machine-based inference is one of the main foreseeable features of Wikidata, and the protocol for that should be a community discussion, not the decision of one person. Emw (talk) 02:23, 28 March 2013 (UTC)
  •  Comment I cannot help but marvel at this statement. I am not trying to write The Singularity here. If F is father of C, then C is child of F. This kind of reasoning does not exactly require a college degree. If you cast your keen eye over Wikidata:List of properties, you can actually see many such connections annotated as "pair" there. I did not decide those pairs, nor did I write them down there. I merely aim to automate the required edits, as it seems that the humans here are too busy making silly arguments to do this manually. If you worry about the actual implementation, I am happy to share the code with you. --Magnus Manske (talk) 09:08, 28 March 2013 (UTC)

• OK, great, so ImplicatorBot will use the 'Pair' column listed in Wikidata:List_of_properties to add logically entailed statements with the inverse functionality properties defined there. Is that right? This is precisely the sort of documentation I'm looking for, which I think should be reasonable to expect especially given the broad scope indicated by this bot's description as something that would "automatically add implicit statements". Could you please describe at a high level how ImplicatorBot would edit Wikidata to reflect any other types of inferences? Regarding the source code, it'd be ideal if it were publicly available. Many thanks, Emw (talk) 03:47, 29 March 2013 (UTC)

I hereby withdraw my request for this bot account. I'm sick and tired of the bickering here, and I don't have either time nor inclination to keep begging. You want to keep Wikidata inconsistent, or prefer to fix it by hand, fine. Code's here (GPL) in case someone else is willing to put up with this BS. Best of luck. --Magnus Manske (talk) 09:16, 2 April 2013 (UTC)

Don't withdraw, please! We need this work. --Ricordisamoa 10:03, 2 April 2013 (UTC)

---

The above discussion is preserved as an archive. Please do not modify it. Subsequent comments should be made in a new section.