About this board

Discussion related to the old Growth team is archived at Talk:Growth/Growth 2014.

Comment encourager les novices à modifier davantage ?

Nemo Le Poisson (talkcontribs)
Trizek (WMF) (talkcontribs)

Merci Nemo ! Les conclusions de cette conversation globale ont été résumées ici et il me semble que tes commentaires rejoignent ceux d'autres personnes.

Reply to "Comment encourager les novices à modifier davantage ?"

Add a link: word exclusion list (per language)?

Gunboz (talkcontribs)

Hi! I wonder if there is an easily accessible ”exclusion” list, which tells the system not to use certain words as ”link targets”?

As an example, in Swedish, my main language, the words ”Lika” and ”Ingå” often appear as suggested links. While Lika is a province in Croatia (=suitable link target), in swedish it also means ”same as” (like in ”like for like”). Needless to say, that this creates quite a few ”false positives”. The same with ”Ingå” (a smaller river =suitable link target), the word however also means ”is included in”.

If I knew where to add words to such a list (if it exists), then I’d be happy doing this myself as I run into them?

KStoller-WMF (talkcontribs)

Thanks @Gunboz for the feedback!

We are evaluating this and several other ideas for improving "add a link" accuracy, while also investigating ways to make these edits easier to review.

What's your experience like with "add a link" tasks? Besides occasional incorrect suggestions, do you feel like this task is working well on Swedish Wikipedia? Do you have any experience patrolling these tasks, and if so: do you have any recommendations for making these edits easier to review?

Gunboz (talkcontribs)

Hi and thanks for reaching out!

  1. I think the task/ feature is excellent: it’s like ”low threshold” editing when I’m too tired or similar to write or proofread outright. It brings me to surprising new subjects, and quite often I also end up fixing some other small stuff after posting the links to an article. And it helps finding quite a few ambiguous articles (pls see my other suggestion on that subject on this page).
  2. Re. patrolling: I have roll-back authority (and have used it on occasion when seeing stupid stuff happening to articles on my watch list), but I have not yet performed any ”systematic patrolling”. I will now check with some more experienced users on svwp, how I could go about doing that (unless you have some ideas or hints), and then get back to yourself in a week or so, reporting on my experience on patrolling these add a link edits, if that would be useful to you and your team?
KStoller-WMF (talkcontribs)

Thanks so much for the feedback, @Gunboz !

  1. That's great to hear!
  2. That would be so helpful, thank you! Since "add a link" makes these small edits so easy to complete, we've had some patrollers mention that it's a burden to review them all. I'm curious to know if this has been an issue on svwp, or if svwp has any other feedback or recommendations for improving "add a link". Thank you!
Gunboz (talkcontribs)

Hi again @KStoller-WMF! I’ve had a look at this. First of all, I don’t know if this has been an issue on svwp (yet), one reason being that the feature was just released there very recently. However, I noticed a couple of (possible) issues with how patrolling these links does work today:

  1. One gets to see ”the whole article” on the left side of the screen, just like with any other edits that have been performed on an article, and then a comment on the right side, saying like ”article xxx has been linked”.
  2. This setup is imho not really helpful: a) one does not need to see the ”whole article” (left side) in this case, as when reviewing an edit of content. Maybe just the sentence with the new link would suffice, possibly the sentende before and/ or after could be helpful in addition, but absolutely not more. This would save a lot of unnecessary scrolling in the linking article. b) On the other hand, on the right side, where today only the link target article name is shown, this is definitely to lean. There is no way to determine if a link to Swan is to the animal or the sailing boat by just seeing that one word ”Swan”. If here instead the first one or two sentences of the target article would be displayed, it would be immediately clear if the new link is suitable or not (in this way I can immediately see the context around the new link on the left side and then the subject where it links to on the right, right next to each other, helping me take a split-second decision re. suitable or not).
  3. The result of such a change/ suggestion would be, that the patroller would be presented a table of linking snippets on the left (text around where the link was inserted) and a clear description of the link target on the right, where each table row represents one new link. This would mean, that the table has two main columns. One could also add a third column on the left side of the table with a hyperlink to the ”linking article”, in case the patroller would like to access it for further overview or editing, although I doubt that this would be used very much in this context. And on the right side there should definitely be a hyperlink to a ”remove link” function, where the insert of the new link could be easily reverted by the patroller, if deemed unsuitable. In all a table with 4 columns.
  4. With such a setup I believe, that the patrolling of these links would cost 1-5 seconds per new link, and I think that I would prefer to get a chunk of 20 to 100 such new links to verify in one go, as opposed to one by one, as with normal content edits. And I think that this would be feasible, since newly added links (especially as in this case, based upon the recommendation of the system) can create considerably less harm, than destructive editing can, so that this patrolling is much less time critical, i.e. one could wait and collect a batch of these added link edits and then let a ”link patroller” sink his/ her teeth into one such batch.
  5. The way the patrolling of this has to be done today, it is imho definitely neither easy nor fun. I believe that this would be completely turned around if based on the suggested table- and batch method.
  6. Needless to say, that I would be happy to take a look at or beta-test such a function, if you were to go ahead and give it a try.
KStoller-WMF (talkcontribs)

Thank you @Gunboz for the extensive feedback! We are still in the early stages of determining how we can improve the patroller experience for these edits, so this is super helpful. I will likely have further questions once we can focus on this more, and I'll be sure to reach out if I have further questions or we have some designs ideas ready for feedback. Thank you!

Gunboz (talkcontribs)

There may also be a next step to this: The other day I had a chat with another user on svwp, @Plumbum208, who actually brought me to you guys in the first place. His analysis, using Petscan was, that the Add a link tool "finds easy links", i.e. links to certain relatively well known subjects, which then end up with a lot of incoming links, possibly leaving other subjects "by the wayside" which might be rarely used and less commongly known, but still (and possibly therefore) would deserve to be more noticed. One use of this insight, after the patrolling issue is handled (btw. svwp does not enforce patrolling, like for example dewp, which might make this issue less severe on svwp), may be to present more than three link suggestion for an article, maybe also in a table with text snippets, then sort these based on the number of incoming links to put the less used on top, or some variant of this idea.

KHarlan (WMF) (talkcontribs)
Reply to "Add a link: word exclusion list (per language)?"

Add a link: Feature idea - „Target article name ambiguous“

Gunboz (talkcontribs)

Hi and thanks for a great feature! Having been using it for a couple of days on Swedish Wikipedia, I have run into the following situation a few times:

  1. The algorithm marks a word and suggests a corresponding article.
  2. Reviewing the suggested article shows that the title is correct, but it still is the wrong target, since it has an ambiguous name without for example a qualifier between parenthesis. A (maybe farfetched) example could be one article „Naples“ as opposed to two articles „Naples (FL)“ and „Naples (Italy)“.
  3. Since this function/ algorithm keeps „finding“ this kind of articles would I like to suggest implementing a means of capturing the valuable insights they give. As it is now, one can only say „No“ to the link suggestion and then tick „Other“ as the reason (unfortunately also without being able to add a comment), and thus the insight is lost.
  4. If, above the checkbox „Other“ on the „Reason“ screen after clicking „No“, there would be a checkbox „Target article name ambiguous“ or similar, followed by two fields suggesting the two new article names, then this information could, for example, be posted as a template entry in the target article, thereby inviting more advanced users to possibly rename the identified article to clear the ambiguity and also maybe create the second one, i.e. the one that „Add a link“ revealed „to be missing“.
  5. In this way the feature would be useful in identifying these ambiguous article names, letting the newbies contribute in a way they now how to (delivering the article names they believe should be there), and then leaving enough bread crumbs for more experienced users to fix the issue.


Trizek (WMF) (talkcontribs)

Hi Gunboz, and thank you for your feedback!

We are collecting the reasons given by users when they click "no", but we aren't collecting "suggested link is too ambiguous" as a reason for rejection. It is a idea we should consider, thank you for suggesting it.

The example you give about Naples is precisely where a human is needed and valued.

It is not possible to make comments for a simple reason: read all of them requires a workforce we don't have. :)

Regarding posting it on the article talk page, I'm not sure to get your point. Do you suggest to have newcomers adding these links no matter what, and then having experienced users to handle them, or do you suggest to have these possible links to be suggested on the talk page?

Gunboz (talkcontribs)

Hi Trizek and thanks for your quick response, and I agree, that this is perfect human-KI-integration, I love it! Sorry if I was being unclear in my post. My suggestion would be (using a fictitious example):

  1. The ”Add a link” system (in an article about ”Volcanoes in Europe”) finds the word Naples and suggests linking it to the Article ”Naples”, which however covers Naples in Florida.
  2. As a user, I now mark this as a ”No”, i.e. ”do not link to the suggested article” (since that would be wrong).
  3. The system now presents me with the ”Reasons” screen.
  4. Now the new, added choice on that screen is ”Target article name ambiguous“
  5. Ticking this box provides me with two string fields, where I can enter my article suggestions: ”Naples (FL)” and ”Naples (Italy)”. The first one would be my new suggested name for the already existing ”Naples” article (whereto I think it should me moved/ renamed). The second one the name for a possible new article (the one that I would have needed for my ”Volcanoes…”-article).
  6. The system could then put a template/ tag at the beginning of the original ”Naples” article (not the talk page) like: {New suggested disambiguation | Disambiguation page = Naples | Current article name suggestion = ”Naples (FL)” | Suggested additional article = ”Naples (Italy)”}.

The template/ tag under 6. could then entice an experienced user to get his/ her hands dirty and implement the necessary disambiguation. This could not be automated. Neither could there be a link entered in the ”Volcanoes…” article at this time, since there simply is no article about ”Naples (Italy)” (yet). However, imho here the ”missing link” is the minor issue, compared to the unresolved ambiguity, which the tool ”just happened” to stumble upon. Through this addition to the ”Add a link” tool, it would start helping find and point out those ambiguities.

Reply to "Add a link: Feature idea - „Target article name ambiguous“"
Iniquity (talkcontribs)

Hello! :) Where can I find out about new strings that need to be translated? I haven't found any news that a new feature pack has arrived (opt out mentorship). And I have some lines either not translated or with poor translation.

@Trizek (WMF), can you help me with it? :)

Jdforrester (WMF) (talkcontribs)
Iniquity (talkcontribs)

Thanks, I also found out about this link a couple of hours ago :)

Personally, it will help me, but if I suddenly leave (busfactor), how will others know that an update has arrived?

Trizek (WMF) (talkcontribs)

I really hope that you won't get hit by a bus, Iniquity! ;)

All strings are located at this address. Only a regular check can help someone who doesn't use watch-translations, I'm afraid.

Iniquity (talkcontribs)

> I really hope that you won't get hit by a bus, Iniquity!;)

I hope so too xD

> All strings are located at this address. Only a regular check can help someone who doesn't use watch-translations, I'm afraid.

Just given that this interface is being shown to newbies and is under active development (i.e. there are new lines) it is very important that it be translated, in my opinion. This affects the engagement and understanding of the interface :( Can we come up with a scheme where new lines would be published?

Trizek (WMF) (talkcontribs)

We plan to have much more regular updates of our newsletter, to inform communities about new releases.

I usually post messages to the mentors' talk pages at the wikis when something big is scheduled, as these users are more likely to help. Would it help if I also inform mentors of important changes that needs to translate a few new strings?

Iniquity (talkcontribs)

Of course! This will be the best solution :)

Trizek (WMF) (talkcontribs)

Noted. :)

Reply to "Translation strings"
Juandev (talkcontribs)

Is there a video about what the newbie see? Recently I figured out that newbies ask questions on something we dont know how it looks like. I.e. newbies receive dialogs we dont know about.

Martin Urbanec (WMF) (talkcontribs)

Ahoj @Juandev! Děkujeme za zájem o funkce týmu Growth i Tvou otázku. Ano, nějaká videa existují. Na české Wikipedii jsem například před nějakou dobou vyvěsil videa ukazující, jak registrace funguje dnes a jak fungovala dříve. Díky tomu je možné změnu v uživatelském zážitku snadno a rychle porovnat. Jsou sice, pravda, trošku staršího data, ale myslím, že pro ukázku jsou vcelku dostačující (ale stejně je zkusím v nejbližší době aktualizovat).

Je také možné vyzkoušet si funkce týmu Growth vlastnoručně. Buď registrací nového uživatelského účtu (čímž získáš přehled o tom, jak cesta nováčka probíhá od A do Z), anebo zaškrtnutím možností „Zobrazit Domovskou stránku nováčka“, „Odkaz na uživatelské jméno vede na Domovskou stránku nováčka“ a „Povolit panel Potřebuji pomoc“ ve svém nastavení. První dvě zaškrtávací políčka jsou na kartě Údaje o uživateli, to poslední na kartě Editace. Podrobnější instrukce najdeš na Help:Growth/Tools/Enable_the_Homepage/cs.

Kdyby tě o funkcích týmu Growth zajímalo cokoli více, dej vědět – rád otázky zodpovím.

S přáním hezkého zbytku dne,

Reply to "What the newbie see"

Can someone answer my question about A/B test?

Lens0021 (talkcontribs)
ChhTJ096 (talkcontribs)


Sunpriat (talkcontribs)

If I want to use suggestions and analytical data from the homepage, is there a way to opt out of the mentor function? Mentors can take participants who did not have a mentor before and intercept a participant by rewriting to themselves. (for example, you can take any account and start stalking it just because of the technical possibility ) The participant can't do anything about it. The mentor can observe the contribution statistics and activity time for an unprecedented long time (even logged actions that are usually not observed by others) that the extension makes to him. Or let's make a trusted WMF account, to which the participants of each wiki could switch when you don't need your statistics to be collected for another and you don't need a chat.

Trizek (WMF) (talkcontribs)

Remove banned accounts from the mentor panel

DarwIn (talkcontribs)

Is it possible to remove or (permanently) hide banned accounts from the mentor panel, since they are absolutely useless there and only add to the clutter?

Trizek (WMF) (talkcontribs)

Hello Darwin

In theory, in Special:MentorDashboard, indefinitely blocked users shouldn't be visible. Which account do you see there so that we can investigate a possible bug?

Thank you!

Reply to "Remove banned accounts from the mentor panel"

Encourage mentors to check mentees contribution + More Feedback

Jcubic (talkcontribs)

This is my opinion about the Growth feature, I've just checked contribution of the last user that asked a question. I've seen that he has some newcomers tasks done, (added wiki links to some pages). And I've seen that often metees ask how long it will take for their edit to be approved (on Polish Wikipedia edits need to be approved to be visible).

The problem I see is that there are not much connection between mentor and mentees. When I sign up for being a mentor (in Polish Wikipedia) they said that I will answer questions maybe once or twice a week. There is MentorDashboard but I don't think that is really useful. I don't check it that often.

But if there was a page that list all my mentees recent edits that, maybe with some filters not sure what they should be, it would be much better if mentor can look at new edits and approve them faster. Maybe it would be even better of the edits are inline with button to approve, so it will be fast and easy for mentor to just check and approve the edits and hide edits that he checked.

Also one more feature that this page should have is thank you link on each edit.

And all this because if someone add his first edit and get it approved and visible faster, and mentor will give him personal thanks. It will increase the number of people that will keep using Wikipedia as editors. It would probably also be good idea to send notification when newcomer edit is approved, this will "force" them to open Wikipedia and see his/her edits. Those can be Preferences set by default, even on Wikipedia that don't require approval for a change to be visible (like in English Wikipedia). Notification for the newcomers that his or her edits was approved, would probably be boost of dopamine for the newcomers and he or she would like to edit more.

Another idea is that it would probably be fun to add some Gamification to newcomers, not sure how. Gamification is the topic that I'm interested and need to give more focus in 2022.

Tacsipacsi (talkcontribs)

For the FlaggedRevs notifications about accepting edits see phab:T54510.

Trizek (WMF) (talkcontribs)

Thank you for sharing your thoughts with us, @Jcubic!

In recent changes, you can filter down your mentees' edits: Then you have access yo the usual tools to validate pending edits.

Implementing Thanks in Recent changes is a Community Wishlist proposal you can vote for (if not already done).

I agree on the fact that quick validation would probably help newcomers to edit more. About thanking users, it has been proven that it helps on retention.

I'm keeping your suggestions in mind for our upcoming thinking regarding how to help mentors. Given the feedback we already received, we will focus on the Mentor dashboard first.

Regarding Gamification, it is something we have considered. We try to pursue with Add a link (available at pl.wp, have you tried it?) and with Add an image.

Jcubic (talkcontribs)
Trizek (WMF) (talkcontribs)

It is not a bug: we have assigned a mentor to everyone for practical reasons. I understand it is confusing; you are not the first person to mention it.

We plan to work on a feature to opt-out mentoring so that experienced users won't have a mentor, I hope we can work on it soon. Edit: It is also possible to filter down your mentees on the mentor dashboard: you can set a maximum number of edits there.

Reply to "Encourage mentors to check mentees contribution + More Feedback"