Talk:Requests for comment/Related Pages

From Meta, a Wikimedia project coordination wiki
Jump to navigation Jump to search

Community support[edit]

Hi. Has anyone ever asked for this feature? Is there a Phabricator Maniphest task in which someone says that a "related pages" feature would be a good idea? Is there a discussion on a Wikimedia wiki where people are asking for this type of "related pages" feature/functionality? --MZMcBride (talk) 23:11, 23 March 2016 (UTC)

Good questions, I'm interested in the answer too. IIRC, this sort of thing started with an ElasticSearch experiment and since then WMF has kept pushing similar things. It would be nice to publish the 2014 results, before starting new experiments. Nemo 07:45, 24 March 2016 (UTC)
I don't think that community support is necessary to start an experiment, but of course at some point it is necessary to check if it is going in the right direction. I guess this RFC fulfills that purpose. I find the feature "mildly" useful, specially considering that most pages have a "see also" section. I have no idea if that was taken into account before developing the feature. Maybe someone can point to the initial discussion to see the rationale?--Micru (talk) 13:03, 24 March 2016 (UTC)
And also, are there statistics about how often those "related pages" links are being clicked?--Micru (talk) 13:05, 24 March 2016 (UTC)
On Nemo and MZMcBride 's points, btw I have replied to MZ's concern on my talk page. Thanks--Melamrawy (WMF) (talk) 21:25, 24 March 2016 (UTC)
Micru both the initial impetus and click-through rates can be found here. Spoiler: the latest is a mobile web click-through of 23% (it is actually going up) and a desktop click-through of 3% (relatively volatile). Jkatz (WMF) (talk) 17:41, 28 March 2016 (UTC)
Experiments seems relevant. --MZMcBride (talk) 03:51, 29 March 2016 (UTC)

FYI: On German Wikipedia there is a poll in preparation, asking the community to explicitly vote for activating that feature, otherwise it should be abandoned. See w:de:Wikipedia:Meinungsbilder/Mehr erfahren. Greetings --PerfektesChaos (talk) 11:51, 27 March 2016 (UTC)

Thank for the headsup, PerfektesChaos. --Melamrawy (WMF) (talk) 13:00, 28 March 2016 (UTC)
German Wikipedia has formally refused introduction of that feature a couple of minutes ago, with 22 Aye : 126 No. Greetings --PerfektesChaos (talk) 19:20, 15 June 2016 (UTC)

Local Wiki posting[edit]

Per discussion with the Reading Team, there is an invitation for the RFC to be brought to other wikis to obtain more thorough community discussion. I've posted it to:

Moved: EN:WP:Related_Pages_extension/RfC discussion is on the Talk page.

Hopefully some other people will jump in to bring it to other wikis. My foreign language ability is approximately nil. Alsee (talk) 09:20, 27 March 2016 (UTC)

Hello! I have 2 suggestions:
  1. This page cannot be translated. If you want to involve more communities, I think it will be better to rebuild this page to allow proper translations and then send a massmessage to the translators asking for their help with that. You can use the Harassment workshop page as an example.
  2. Have you considered sending a mass message to all the village pumps? I think those can be marked for translation too. I can help with the spanish on.
In any case, I will post a link to this page in the Spanish Wikipedia's VP.
Regards, Lsanabria (talk) 13:10, 31 March 2016 (UTC)
This is the note I left on eswiki: w:es:Wikipedia:Café/Archivo/Miscelánea/Actual#Función Beta: Páginas relacionadas --Lsanabria (talk) 13:41, 31 March 2016 (UTC)

Recommendations for moving forward[edit]

Hi. I posted my thoughts on moving forward here: mailarchive:wikitech-l/2016-April/085209.html. --MZMcBride (talk) 22:10, 3 April 2016 (UTC)

Which reminds me, some (non-Wikimedia?) wiki might want this feature for dead-end pages. Would be nice to have a global for that. Nemo 07:38, 4 April 2016 (UTC)

overlap with see also[edit]

In response to: http://www.gossamer-threads.com/lists/wiki/wikitech/697996#697996

Thanks, Vituzzu! This is all great feedback and exactly what we are looking to get from the RFC. To be clearer to others who may have missed it, the RFC is a request for comments about this test feature, not a proposal to roll-out the feature.

I appreciate your concern around redundancy. I think that is one of the biggest, most legitimate concerns about the feature (the other being how to deal with bad results) and, honestly, one of the most obvious outcomes we might have avoided had we started with a consultation rather than by building. As Jdlrobson mentioned here, http://www.gossamer-threads.com/lists/wiki/wikitech/697997#697997, there is some active discussion around how we can improve on this and many of your comments fit well within that scope. As to better images, this is also something we are working on, primarily here: https://phabricator.wikimedia.org/T91683

Despite not being perfect, we felt that related pages add unique value on top of the "see also section" because it offers the reader a limited selection compared to see also and sits at the bottom of the page, where it does not distract from the article, because a user reaching the bottom of the page has finished reading the article and is theoretically looking for more content. Notably, we see that mobile users reach the bottom of the article less frequently, but when they DO and they see related pages, they are much happier to see it (as indicated by click-through). It might be that this feature is better suited to mobile.

As I wrote on the project page, I'd like to identify if the overlap with "See also" mean's you would rather not have the feature or if we feel that there is still positive value. We are trying to identify the order of value here for our readers: no related pages < related pages with no new features < related pages more customizable < related pages features better synchronized with/eaten by "see also" (is the above order correct? I imagine that community members might want to swap at least the first 2 with each other...but I don't know for sure)

Thoughts?

Jkatz (WMF), 18:44, 4 April 2016‎ (UTC)

How are they selected?[edit]

Are relevant pages selected by pageviews or the number of incoming wikilinks, or both, or something else? EllenCT (talk) 21:58, 5 April 2016 (UTC)

EllenCT, I don't deal with this product myself, but I believe that it's the first three pages that you would get if you typed the current page's name into the Search box (excluding the current page, since there's no point in linking back to the page you're on). Whatamidoing (WMF) (talk) 19:04, 7 April 2016 (UTC)
Thanks. Maybe there is a link to the code that does that somewhere. EllenCT (talk) 23:44, 8 April 2016 (UTC)
Hello EllenCT, in fact there is a link to how this works here thanks!--Melamrawy (WMF) (talk) 12:03, 9 April 2016 (UTC)
Thank you, @Melamrawy (WMF): At [1] does the local implementation of the doc-boost() function involve pageviews, inbound wikilinks, both, or something else? EllenCT (talk) 23:27, 9 April 2016 (UTC)
EllenCT Thanks for caring! My understanding is that boost takes into account both inbound wikilinks and whether or not an article has a particular set of templates (on Eng Wikipedia- the only template in the list is featured article). The current issue I have with the implemtation is that those two represent 2/3rds of the weight in making a decision to show an article, with text in common only making up 1/3. This is the kind of thing we can fix and we have a ticket for it: T128822, but it is an effort and I only want to do it if there is a chance that the feature would be rolled out. Now to complain and ask for advice: I am currently having a hard time gauging whether the proposed solutions to community concerns in the RFC would address the concerns. If we moved forward as proposed, would people support a roll-out of related pages on mobile web or desktop?. Any ideas? Jkatz (WMF) (talk) 19:32, 14 April 2016 (UTC)
I don't think tweaking the algorithms would matter much on the basic issue. There are reasons to support the concept, and reasons to oppose the concept, but algorithm tweaks aren't likely to shift the basic rationales.
I'm disappointed at the low response rate it's been getting (at least on EnWiki). I didn't anticipate how much of a impact it would have to deviate so far from the usual RFC style. We specifically steered away from determining consensus about rolling it out. Even so, I should have stuck with more conventional Support/Oppose sections, perhaps encouraging people towards something like an "I would support if" section in the middle. I just added a new link to the RFC on the Village Pump Proposals page which should bring in more attention. Alsee (talk) 12:46, 15 April 2016 (UTC)
@Alsee and Jkatz (WMF):: Just a pointer: As I mentioned in February in Mediawiki talk, I'm in the process of setting up a so-called "Meinungsbild" in German-language Wikipedia at de:Wikipedia:Meinungsbilder/Mehr erfahren. Meinungsbilder are German Wikipedia's more formalized community consensus finding process; somewhat similar to English-language Wikipedia's RfCs, but with more emphasis on voting. The Meinungsbild (MB) process is multi-layered, so to speak; in order to hold a valid MB, one has first to gather ten supporters acknowledging that the MB is sufficiently prepared and can be held. That often takes some time (how quickly these ten support votes are gathered also depends upon general community interest, which seems to be a bit low in this case). Then, a date for starting the MB is fixed (not earlier than a week after gathering ten supporters); and then, MBs consist of two parts: A vote on formal acceptance (like "do you agree with this MB at all or reject it as a whole?"), and one on the actual proposal (in this case, "if the feature leaves beta and would be ready to deploy, do you support it being activated for German-language Wikipedia?"). The usual running time for a MB is two weeks after start. Currently, the MB "Mehr erfahren" (learn more) is still in the "gather ten supporters" state. So, in German-language Wikipedia, we will have a result on this question in a few weeks at the earliest, depending on MB support. Gestumblindi (talk) 18:23, 19 April 2016 (UTC)
@Jkatz (WMF): I would certainly add pageviews from analytics into the boost, refreshed weekly. The number of inbound wikileaks is only partly correlated, and won't change as much in response to current events. EllenCT (talk) 23:40, 18 April 2016 (UTC)
Thanks, @EllenCT:. I apologize--just seeing this now. This is an interesting idea. If we want to use popularity, that is a great idea and we have a new pageviews API to help us do that. However, we were thinking of abandoning popularity altogether, or at least diminishing it significantly. It seems there is community interest value in surfacing rarer articles that a user might not find otherwise. What do you think? Jkatz (WMF) (talk) 00:26, 21 May 2016 (UTC)
@Jkatz (WMF): can we please see a list of what those suggestions of rarer articles would look like on a few dozen example articles before that goes live? I'm pretty sure you want to suggest related popular articles from rare articles, but there isn't any reason you can't still do that if you suggest rarer articles from very popular ones, is there? EllenCT (talk) 00:56, 21 May 2016 (UTC)
@EllenCT: As to your first question, I think that is totally reasonable and we can and will publish a sample before making any change. It will post on the phabricator ticket - on the related pages project page. I just added that requirement to the "task" in phabricator: https://phabricator.wikimedia.org/T128822. If you have a list of sample pages you would like us to try against, please let me know. Otherwise, I have about 10 pages that I look at whose results seem bad to me and the related pages talk page community members have identified another 20 or so. Unfortunately, I could not follow your second question--I didn't think we should suggest popular articles from rare ones: is there a reason you think we should do that? I think it would be hard if we tried to send people on rare articles to popular ones and people on popular articles to rare ones, so want to gauge the value betterJkatz (WMF) (talk) 17:46, 21 May 2016 (UTC)
@Jkatz (WMF): you are already suggesting popular articles from rare ones. I like User:Gestumblindi's that you should make a "See also helper" link that people can use while editing to review such suggestions and decide which ones to include manually. How do you feel about that? EllenCT (talk) 20:17, 21 May 2016 (UTC)

┌─────────────────────────────────┘
@EllenCT: "you are already suggesting popular articles from rare ones", yes, and the ticket I link to above is specifically to change that to suggest less popular ones. It feels like we are talking past each other a bit, so I will try to sum up what I see the issue is: If you think we shouldn't be suggesting rarer articles, please let me know, otherwise I think we're on the same page. Regarding the "see also" suggester, I think that is a valid suggestion. What I said to Gestumblindi below is that such proposals are"valid and interesting, but it feels like they are not necessarily wanted, but more of a compromise. This being the case, I honestly, do not think the feature merits the kind of reworking that would require." Does my understanding fit reality? In other words, do you feel like this 'suggested see more', is in fact, a desired feature that you would like the team to spend ~6 months-9 months on, or is this more of a "if the WMF really wants this on desktop, here is how you should do it"? Jkatz (WMF) (talk) 18:34, 22 May 2016 (UTC)

@Jkatz (WMF): how long would it take to make a two-column list of 20 suggestions each taken from popular and unpopular articles, linked from the &action=edit page as a target='_blank' new tab? EllenCT (talk) 17:04, 23 May 2016 (UTC)
@EllenCT: Well--everything here has a cost both in terms of consultation, rollout and maintenance, but I don't think it is a hard thing to produce actually, given that it is simply an untransformed return from an API. That being said, the scope for these things always grows as you dig into the details: some necessary, some unecessary. Why? to my earlier question, is this something you feel you and other contributors would benefit from?Jkatz (WMF) (talk) 21:33, 25 May 2016 (UTC)
@Jkatz (WMF): yes, I absolutely think it would be hugely useful to have lists of both more and less popular articles. If you need a longer project, how about have checkboxes for editors to recommend which they think would be most usefully related instead of just lists? EllenCT (talk) 21:55, 25 May 2016 (UTC)
@EllenCT: Okay, thank you for clarifying! As to having a longer project, I am trying to shrink the project! Looking at the other things we think are important, it will be hard to schedule this as is for my team, but it is good to know you feel it adds value (as opposed to being a compromise you can live with). I actually think that what you're proposing might be very interesting to the editing team, whose focus is on giving editors better tools. @Jdforrester (WMF): might be interested in particular. Thank you for working this out with me and lookout for further announcements on the future of related pages! Jkatz (WMF) (talk) 19:11, 29 May 2016 (UTC)
@Jkatz (WMF): As I recently mentioned at Mediawiki, inbound wikilinks may also come from navigation templates such as "members of <some parliament>" that are present in many articles but not necessarily mean high prominence. - I think this is a bit of a dilemma: You are, understandably, reluctant in improving the feature as you don't know whether the Wikipedia community (individual language communities) might in the end, despite all improvements, oppose the feature on a more fundamental level (i.e. "we don't want automated links, we dont want the WMF interfering with our manual content curation"). On the other hand, community members that might possibly support the feature probably feel reluctant to do so if they don't know how the feature will ultimately work. Myself, I'm not quite sure what to think. Currently, I see the feature rather negatively and think I would support it only if it were reworked into something like a semi-automated "see also helper": Let it generate a suggestion for a (nicely designed) "see also" section in articles that haven't one yet. But this section would only become active (visible to the public) after manual confirmation, with the possibility of easily changing the suggested links. So, if we think "no, that article doesn't need a see also section at all": just don't activate the suggestion. If we think it's a helpful suggestion: activate it. In that way, as mere suggestions, I think it could gather wider acceptance in the community. - Despite my own views, by the way, I'm of course trying to design German-language Wikipedia's MB (described above) as neutrally as possible, with all reasonable points pro and contra that are brought up. Gestumblindi (talk) 18:48, 19 April 2016 (UTC)
Oh, and I've now added a link to this RfC to the Meinungsbild draft in German-language WP, so people visiting that page maybe chime in here as well. Gestumblindi (talk) 19:15, 19 April 2016 (UTC)
Thanks, @Gestumblindi:!! I apologize. I am just seeing this now. I really, really appreciate you working with German Wikipedia on this, despite your valid reservations. As you can see from the RFC page, we think that the performance on desktop is low enough and the issues raised are serious enough, that we are thinking the next stage is to probably undeploy from desktop beta. The kinds of alternative visions you and others are putting forward are valid and interesting, but it feels like they are not necessarily wanted, but more of a compromise. This being the case, I honestly, do not think the feature merits the kind of reworking that would require. However, the performance (as measured by sustained reader click behavior) is 5x as high on mobile and there seems to be less concern from community members about it. So we are instead looking to rollout on mobile web only, but specifically not on German Wikipedia (unless I hear otherwise), pending two blockers: 1. improve loading time (so there isn't a lag), 2. improving the algorithm by removing the importance of popularity to help readers find pages they wouldn't necessarily find otherwise. Here is the ticket that shows the blocking tasks: https://phabricator.wikimedia.org/T135030. Any updates from the 'Meinungsbilder'? Jkatz (WMF) (talk) 00:26, 21 May 2016 (UTC)
@Jkatz (WMF): Thanks for the reply! As the Meinungsbild "Mehr erfahren" has now sufficient support, I fixed its start date on June 1. It will run for two weeks, so we can expect a result on June 15. I also added a bit based on the information of your reply here (that the feature will probably not be deployed for desktop, but for mobile web, yet not on German Wikipedia unless its community wishes for it). So, we will have a clearer picture of the German-language Wikipedia's community views on the feature in June. Gestumblindi (talk) 00:57, 21 May 2016 (UTC)
@Gestumblindi: Excellent, thank you! Jkatz (WMF) (talk) 17:46, 21 May 2016 (UTC)
@Jkatz (WMF): Probably you have seen the outcome of de:Wikipedia:Meinungsbilder/Mehr erfahren? It's a clear rejection, only 14,9 % of 148 participants supported enabling the feature. Gestumblindi (talk) 23:21, 3 July 2016 (UTC)

Re: Mobile-only[edit]

It doesn't make sense for mobile to be special here, I really don't understand [2]. Nemo 10:02, 15 May 2016 (UTC)

Perhaps it's because navboxes are invisible on Mobile, but not on desktop. This provides a path out of the article on Mobile, just like navboxes provide a path to other articles on desktop.
The value of this feature probably depends upon the development of the local wiki. Readers of a small Wikipedia would probably get more value out of this than readers on a large, well-maintained Wikipedia. Whatamidoing (WMF) (talk) 16:48, 18 May 2016 (UTC)
Hi Nemo, there is nothing special about mobile here, but we couldn't get clear consensus on desktop, if you are interested in pushing this forward, or helping us find in-depth feedback, would you like to start a discussion on Italian Wikipedia? We can enable it there, if there is consensus, and other Wikipedias can apply a similar process. Let me know if you would like to proceed this way. Thanks! --Melamrawy (WMF) (talk) 19:50, 19 May 2016 (UTC)
You didn't get consensus on the desktop but did you get it for mobile? I never saw a RFC for mobile. --Lsanabria (talk) 16:04, 16 June 2016 (UTC)
My guess is that on mobile it's easier to intentionally or unintentionally hit large pictures instead of small text links. On desktop less skill to follow a link is required. --87.148.72.140 17:08, 4 June 2016 (UTC)

mw:Wikimedia Apps/Feeds[edit]

There is also this thing, which seems related. The idea is not yet clear and we don't know at what stage it is; hopefully there will be more information soon. --Nemo 11:39, 27 May 2016 (UTC)