Wikipedia talk:WikiProject Intertranswiki/Archive 5

This page is an archive of past discussions. Do not edit the contents of this page. If you wish to start a new discussion or revive an old one, please do so on the current talk page.

New Research: Characterizing Existing Practices for Identifying and Mitigating Knowledge Gaps

Latest comment: 3 years ago1 comment1 person in discussion

Hi All,

My name is Jim Maddock, and I’m with a researcher at Northwestern University working on identifying missing content on Wikipedia. As a first step, we want to talk to members of Wikipedia’s editor community to better understand how editors currently identify and add missing content. Participants must be Wikipedia editors who speak English and will be compensated for their time.

For more details about our project, please refer to our project meta page. If you are interested in participating, please fill out this screener and consent form. Additionally, feel free to reach out to me at maddock@u.northwestern.edu if you have any thoughts and suggestions. Thanks!

Study Information

Study Title: Characterizing Existing Practices for Identifying and Mitigating Knowledge Gaps

PI: Darren Gergle

IRB Study #: STU00212033

Cheers, Jmads-nu (talk) 16:18, 16 July 2020 (UTC)

Call to action!

Latest comment: 4 years ago1 comment1 person in discussion

pls see Wikipedia:WikiProject COVID-19/Translation Task Force.--Moxy 🍁 15:21, 24 March 2020 (UTC)

Missing featured content

Latest comment: 3 years ago19 comments3 people in discussion

Hi, members of WikiProject Intertranswiki! It's WikiMacaroons. I have recently programmed a python script that searches through Featured/GA articles on other language Wikipedias, and checks if they exist on enwiki. The index can be found HERE. Keep in mind that the translation of article names uses Google Translate, so some of these articles may exist under different names on enwiki. It also lists the other categories the page is in, so you can search for subjects you're interested in. Anyway, hope this is useful. Thanks, Wiki Macaroons^Cinnamon? 18:13, 7 September 2020 (UTC)

Hi, I think this would make more sense to do with a Wikidata query. I'm sure someone there could help you set up a search to identify all FAs in a given language missing English-language equivalents. I opened up a number of the Spanish FAs, and all of them had corresponding English articles under titles other than that generated by the machine translation. Calliopejen1 (talk) 00:00, 12 September 2020 (UTC)

@Calliopejen1: Thanks for your response. Someone else suggested that I identify the article's wikidata entry, but I couldn't figure out the python logistics. I may have to use JS next time. Wiki Macaroons^Cinnamon? 08:06, 12 September 2020 (UTC)

Thank you.

Your featured articles list looks like a useful filter. Thank you for it. I wonder if there's a further filter that would separate out the entries in wiki-de and wiki-es (and ... and ....) that have sufficient accessible usable sources... No, I think for now that would probably need to be done by hand still. But maybe one day ....

As you may have spotted, right now there's only one person translating wiki-entries into English through the project transwiki. It's me. We started off with great wiki-fanfare in around 2015 but after about a year I was the only one left, so I started doing the monthly archiving bit etc. for myself. Several of the people who used to work through this "project" are still producing excellent translations, selected according to their own priorities. For all I know one or two of them many have spotted your message (above) and clicked on your list HERE. But self-evidently, adding your own batches for translation on a monthly basis and then ticking them off in the way this project was set up in 2015 isn't for everyone. I carry on doing it because (1) I'm a creature of habit and (2) doing so maybe encourages other folks to translate wiki entries into wiki-en. But if you really want to flag up your list to people likely to use it, maybe you should distribute it to the individual talk pages of the folks who along the way have signed up for this project, and they can maybe use it for selecting articles for translation regardless of whether or not they want to go through the admin stuff set up on this project page.

The other thing that was very noticeable back in 2015 when there were 10 or 20 working through this project page was that we were all great at recommending what other folks should translate, and occasionally I followed up on one or two of those suggestions, but we generally very much preferred most of the time to make our own selections. I guess we all have our own interests. For myself I already have "my" list of around 1500 red links with possibly usable / translatable / adaptable articles in other language versions. If I were to translate them all it would most likely take me longer than I expact to hang around on planet earth. Especially as the list is still growing.

(Though I am getting better at checking carefully for available sources before starting out on another one. Sometimes the more I look the more I think the sources I had thought sufficient ... I don't like so much....)

Interestingly, when a wiki-comrade launched this project in 2015, it was already described as a re-launch. I think it had already been launched before but then fizzled out, as it would have done again this time if I hadn't kept on with it. I guess wiki-world is peopled by folks who like to organise themselves in their own ways. And of course all our brains are differently wired!

Anyway, if you would yourself like to use the lists you are generating to select wiki-entries that you yourself will translate, that would be fantastic. Whether or not you organise yourself through the admin modalities set out by user:Dr B for this "project" back in 2015, or applying your own administrative modalities is really not what matters, I think.

Success. Be well. Charles01 (talk) 08:57, 12 September 2020 (UTC)

(later)

I just took a slightly longer look at one of those lists. You may have noticed. For me these lists hold out the possiblity of finding the best wiki entries that deserve to be translated into wiki-en.

I clicked on maybe five potted biographical entries (as I hoped) in need of translation. But I found all five had already "been done". They appeared on the list, as far as I can tell, because they had not been done with the article title that you and / or the robotic programme had anticipated. I've no idea whether a workaround will be easy/tricky/impossible. But I thought it might at least be worthwhile to flag up an issue that arose when I tried to use the list as (I infer) you had inten ded for it to be used.

You'll see I changed the entries in question. That highlights where the issues arose. But you should not, of course, hestitate to reverse my "edits" if they get in the way of your way of doing stuff.

I still think what you are doing here is seriously worth doing. But obviously if, for the "customer", a robot programme costs more in terms of (human) time than it saves, further work is needed. The world is full of computer enthusiasts who think that man exists to serve the computer, and that may very well be where we are all headed, if World War III and/or terminal plague don't get us first. But I am still old enough to think the computer exists to serve man. And for the avoidance of doubt woman and non-binary. Please infer no disrespect and sorry if you (with any luck very briefly) choke over your coffee on reading this.

Still think it was worth reporting.

Success. Be well Charles01 (talk) 09:04, 18 September 2020 (UTC)

Charles01, Sorry, I only just saw this. I don't find this rude at all, it's in fact very helpful. In the next version, I intend to use wikidata to track whether articles are connected to items. Again, your edits to the lists were great, thank you. I can imagine that was a pretty menial task to do, but useful for the "customer" ultimately. I will try to rework the code in javascript so it can check definitively whether an article exists.

P.S: In the next version do you think I should list how many references the article has? Just a thought. Thanks again, Wiki Macaroons^Cinnamon? 19:20, 5 October 2020 (UTC)

I'm afraid I've forgotten what I wrote before. Somewhere between the short-term memory and the long-term memory there's a chasm. And it (the chasm) grows! But looking again at what I wrote - for that matter also at what you wrote - I think it all more or less still makes sense.

Listing how many references a wiki-es or wiki-de entry has would certainly be a "nice-to-have" add-on, though it necessarily begs more questions. Do you mean just those coded as inline citations? Or do you mean those sources listed under "weblinks" in respect of which you have to look at the link in order to know which bit of the wiki-entry it refers to? With a lot of wiki-de entries there are more useful sources listed under "weblinks" than there are under "references" or "sources", especially for those of us - generally includes me - restricted to sources we can call up online. But do you, with this, just mean the references that can be accessed online, or do you also mean the ones for which - if I have time and inclination - I can drive up to the university library, renew my reader-ticket, and wait forty minutes while they dig whatever it is out of the stores? (They used to do wonderful choclolate cake in the tea room for while you were waiting, but that was >forty years ago, and these days I don't allow myself chocolate cake. Or even - while the plague is raging at current levels - tea rooms.) There's no right or wrong answer on any of that, of course. Just saying (writing) .... And what you cannot begin to test by algorythm is the difference between references that I would find (1) useful and (2) accessible and the ones which I wouldn't. So ... thinking as I go along, a "nice to have" but probably more useful in theory, and if we don't think too hard about it, than it might prove in practice.

Soooo .... how would I be hoping to see it? A list of non-wiki-en wiki entries - maybe sorted into separate categories according to whether they are (1) biographical about people, or (2) about places (3) about types of insect... etc etc multiplied by about 1,000 (though the place to start is presumably just with the top five or ten). That's something a computer can easily do. And there already is a lot of cartegorisation in wikpedia, so you wouldn't need to reinvent the categories into which you sort things. Just decide which 100 or so of the 10,000 "category possibilities" are most likely to be useful for what you are trying to accomplish. I've inadvertently given you my own top 2 already, but of course we'd all have our own preferrd categories / sets. Then - for myself - I would like to end up with a list of - say - 100 biographical entries in German/French/Dutch/Italian wikis for each of which I can click in turn on the relevant entry because there is, till now, no wiki-en equivlant. If they all have good article status in their own wiki version that can bea useful preliminary filter, and if there are enough, one could probably do the first few lists of 100 with nothing but entries adorned with "good article" flags. After that just A-class? Or A/B? But still I have my own human judgements to make. (1) Do I sufficiently understand the German language text to attempt a usable translation / improvement / elaboration (delete those that do not apply according to circumstances)? (2) Do I find it sufficiently interesting? (3) Regardless of wiki-status, do I think it's a good enough article to be worth checking out in more depth? (4) Are the sources - enough of them - good enough that I can satisfy myself it's not all made up? (5) Are there additional accessible sources I can use to find out more about the bits that I find particularly interesting? Self evidently my list of criteria will be different from yours, even with just five. And I will sequence them differently each day of the week. But from the point of view of what I understand of your vision with this, you should not expect to try and replace everything I can do with software. We've got google-translate for doing that already. (And the results are still often hilarious or dangerous or both. Computer as tool, please: not as boss.)

Ho hum.

Success. Charles01 (talk) 08:29, 6 October 2020 (UTC)

@Charles01: Firstly: My original plan to count references was to find out how many times "<ref>" appeared on the page, but if what you say about dewiki and weblinks is true, I could also find all the instances of "https://" and "http://" on the page.

Secondly, you'll notice in the lists I've currently made, I have a list of "Additional categories", translated from German/Spanish. There could be a better way of doing this, but those are in place so you can "Ctrl+F" for topics you may be interested in.

Also, I think I'll stick with GAs and FAs for now, as other language Wikipedias have different standards for what qualifies, so with those flags you can pretty safely assure they'll be of high quality.

P.S: I do like myself a good chocolate cake, too. Best, Wiki Macaroons^Cinnamon? 17:12, 6 October 2020 (UTC)

Yes, the "https://" and "http://" solution doesn't get round the "opaque templates" problem (see my comment a couple of sections below in answer to someone else's timely conundrum), but probably it'lll pick up more relevant info from the page being interrogated than "<ref>" or similar. Then again, as ever, when faced with the salesmanish question "do you prefer the green socks or the grey ones?", my immediate reaction is "Please, neither" (if I do not need socks) or "Please, both" (provided the quality and price are ok)!

Please ignore that stoopid digression about socks: doesn't take us anywhere usefil.

Success Charles01 (talk) 07:56, 7 October 2020 (UTC)

Charles01, Ok, here's another solution: I could find the html content of the page and find all the links that don't begin with "https://de.wikipedia.org"...

I think this is the part where I say "Success." Wiki Macaroons^Cinnamon? 17:02, 9 October 2020 (UTC)

And then include a little column in the results table showing the number of links that haven't been excluded for that reason (or for other reasons we dodn;t think of? That could indeed be a useful add-on.

My blind spot about where templates get their instructions from is too complete for me to have (anything more than) the faintest idea what effect this might have. But I can't see how it shouldn't work (nor how it should ...).

I wonder if there's anything else I should have thought if. If there is, I don't know what it is. Yet.

Good weekend Best wishes Charles01 (talk) 07:46, 10 October 2020 (UTC)

Charles01, Well, are there any other considerations you make when choosing whether to translate an article? Success. Wiki Macaroons^Cinnamon? 10:17, 11 October 2020 (UTC)

Nothing in the abstract. I could list them as bullet points, but I think my own conscious criteria are buried - not too deeply - in the discussion above. And you can pull out the bullet points as well as I can. There might, of course, be other criteria of which I am not consciously aware. But if you look at my edit history you can see where I personally tend to land up. Not that you should be planning this thing for one person (unless yourself!)

The first showstopper at the moment is all the false positives on the list so far to which I alluded before. If I got through five or ten listed entries on German wiki, click on each one, and find that for each in succession, only to find that there is already a long-standing wiki entry on English wiki which, for whatever reason the gismo didn't spot, then I am, frankly, unlikely to be in any hurry to rush back and go through the next five or ten. But once that has been sorted, then I might come back to it and so, more importantly, might others. But the first challenge is to make sure it "does what it says on the tin", to quote an irritating advertising slogan which for some reason comes to mind. Bells and whistles are great, but they're not the place to start.

It is, of course, unavoidable that if I or anyone else started using it on a regular basis, we would in due course think of other things which we'll wish we'd thought of and better understood earlier. But that's just like any other day in the life of a systems analyst!

Be well. Charles01 (talk) 13:47, 13 October 2020 (UTC)

Charles01, I'd better get to work on that then! Thank you for all your suggestions, Wiki Macaroons^Cinnamon? 17:44, 16 October 2020 (UTC)

@Charles01: The first experiment was a success! I tested it out with French FAs, and intend to try German next. View the table here. Feedback appreciated! Wiki Macaroons^Cinnamon? 07:17, 25 October 2020 (UTC)

Ok, German one is done too. See here. Wiki Macaroons^Cinnamon? 10:55, 25 October 2020 (UTC)

First reaction

Thank you. I got one false positive (which I highlighted by changing it) and a second (which I failed to highlight -sorry) where the affair/scandal which you had picked out as a French language wiki featured article and which is covered - for my taste adequately covered, but that's a matter for individual judgment - in the article on the town village where whatever it was happened.

But basically the French and German lists worked well for me and can provide a good alternative route for picking out and / or prioritising entries for translation into anglophone wikipedia.

I had not realised - and maybe I should not extrapolate this as my generalised impression from the small sample that I landed on but here goes anyhow - just how looooong many featured articles can be. I wondered if it was because French readers and German readers have longer attention spans than anglophones. But actually there are plenty of featured articles in English language wikipedia - especially on North American topics - that are loooong for my taste. But the joy of wikipedia is that we are not under cash contraints when it comes to the price of computer-memory (even if Ms Maher seems to have me on her list of people who might very occasionally donate small amounts of money... from time to time). Nor do we have editors / agents / publishers insisting that we don't go above 5,000 or 80,000 or 250,000 or, or ... words per topic. One is almost invited to go on for toooo long. But there might be readers who wanted that level of detail. And of course, sometimes I might be one of them.

Further thoughts arising

These are personal to the way I could and very probably would use your programme, based on what I (currently) do. They are not necessarily a comment for more general application.

For me personally, a Dutch-language equivalent and an Italian-language equivalant could also be helpful. And a lot of people in England and North America might be interested in Spanish version if they knew about it. However, I don't even know if Italian, Spanish and Dutch wikipedias have their own equivalents of featured articles. So if you find that I know not what I ask.... it's true.

After an initial very brief browse, I simply went through picking out biographical entries. Did your programme pick out all featured articles that are biographical articles, or just the ones in one or two of the categories that I personally seem to have focused on lately (without necessarily meaning to)? I suppose historians get lots of hits because they crop up in lots of source notes. But the architects and the lawyers ...?

Thanks again. Be well Charles01 (talk) 10:54, 26 October 2020 (UTC)

Charles01, the script was intended to go through the featured articles category... though it might've not picked up all of them. I'm going to pretend that I'm psychic and predicted the categories you favour. I may try a couple of other languages. Success. Wiki Macaroons^Cinnamon? 17:24, 28 October 2020 (UTC)

Not psychic? But I was taught that the computer only answers the question you asked him? Maybe that's out of date. It is, in that case, interesting that different wiki-versions seem to find different kinds of notable entties worth working up to a featured article level. Then again, maybe I shouldn't over-interpret the priorities of half a dozen or so keen contributors of a certain type...working on each of the different language versions.

At the moment the lists look as though they may be in a temporary location. If you are thinking to put them somewhere else more permanently, please advise me when and where. If you want to make them sub-pages of the intertranswiki page, that would make sense. But there could be better places. I don't have a clear enough knowledge of the overall wiki-architecture (if any) to have a very clear opinion on that. Same considerations apply to any future language versions you might do. After that ... these things move quite slowly, but it might be helpful if you were to schedule a yearly or two-yearly re-run / update? (Before that, of course, you may have thought of improvements. And / or I and / or A.N.Other might have come up with and shared a few more reactive thoughts.) Success Charles01 (talk) 11:21, 29 October 2020 (UTC)

Teahouse question for this project

Latest comment: 3 years ago2 comments2 people in discussion

You are invited to join the discussion at Wikipedia:Teahouse § Copy-and-paste non-English pages problem. {{u|Sdkb}} ^talk 06:03, 7 October 2020 (UTC)

I've had this problem too, but I do not know the answer. Templates seem to reference something in another corner of the wiki-woods - maybe even far outside ther wiki-universe - but I do not know the key to finding out what or where it is. Sometimes I work round it by somply copying the "outcome" of the template's efforts - ie on the wiki-page, without attempting to understand what is going on with the templates on the edit page. That can be a reasonable work-around for copying the salient details of a reference link. But where - I think it's on Italian or maybe French wiki - the info box references a template for basic data, then this workaround doesn't really work. Quicker just to copy the lines in question afresh.

Success. Charles01 (talk) 07:56, 7 October 2020 (UTC)

Rfc on machine translation

Latest comment: 2 years ago5 comments5 people in discussion

In 2016 this content translation tool was completely blocked from machine translation into English as a first step in translation. As machine translation has improved since then should it now be completely or partially unblocked? Chidgk1 (talk) 19:00, 16 January 2021 (UTC)

Comment @Chidgk1: has it? By how much? Can you give us some data or examples? – Finnusertop (talk ⋅ contribs) 11:58, 21 January 2021 (UTC)

@Finnusertop: History_of_machine_translation#2000s only goes up to the 2000s so I have tagged it to ask for it to be updated for the 2010s. According to a 2020 study: "in recent years" there has been a "dramatic improvement of translation quality" because "the field of MT has shifted to the use of deep-learning neural-based methods". ^[1]

References

^ https://www.nature.com/articles/s41467-020-18073-9

Strong support This should be available to extended confirmed editors only and it should be made clear that use without cleanup and proofreading will result in a ban from using the tool. For responsible editors, it makes the process of article translation and attribution vastly easier, and is much better than what I saw sometimes patrolling new articles where an editor would copy and paste a foreign language article into Google Translate and pass it off as their own work on enwiki. (t · c) buidhe 21:36, 10 February 2021 (UTC)
Support I expect that people wish to avoid poor user experiences following their use of the translation tool, then getting the feedback that their submission is unacceptable. I am less concerned about the quality of translation than I am with users having a positive experience trying to use the tool. Is there past evidence that the tool caused problems? I expect that there were few or no problems, and that having another period of testing right now would be fine. I expect that there would be less than 1000 attempts to use the tool in a year, and that the usual quality control processes would catch unacceptable use easily, and that much more good than bad would come from it. Blue Rasberry (talk) 21:49, 15 February 2021 (UTC)
Support, with reservations. The problem stems from those who, as Buidhe mentioned, simply paste machine translations into the article. I used to try and stem the tide, such as at PNT (see WP:PNT/by user, and this list, for example) but it was more like a tsunami and I eventually gave up; all of the articles in that list remain available, afaik, and have never been checked for accuracy. The fact that machine translation is far better now than it used to be, is a mixed blessing: the wording that comes out of it is no longer laughably poor English, and sometimes is excellent grammatically speaking.

The downside is, that the result may fail to accurately represent the original meaning, and sometimes flips the sense of a sentence 180 degrees—but since the English is perfect or nearly so, a monolingual cannot see that. The template {{Hidden translation}} is available for this, but never caught on. They key to adding translations, whether manually translated or facilitated by MT, is that the editor who publishes the result must be knowledgeable in the source language, so they can spot the errors and fix them. Speaking perfect English is simply not good enough. I certainly agree that those with competence in another language should not be forbidden from using MT to facilitate a translation; if nothing else, it saves a ton of typing, even if I don't always agree with their wording choice (and especially, their syntax order). Mathglot (talk) 22:09, 27 February 2022 (UTC)

Help from a translator

Latest comment: 3 years ago3 comments2 people in discussion

Hello. I need help translating the article Fabrizio Romano into Italian. Could someone assist me in this process? Thank you. Paul Vaurie (talk) 00:24, 1 March 2021 (UTC)

Note: I already started at it:Utente:Paul Vaurie/Sandbox Paul Vaurie (talk) 00:26, 1 March 2021 (UTC)

I don't think anyone reads this page except me. And I don't look at it very often. Also, my Italian is nothing close to good enough for your purposes. If you want someone to help with translatiing into Italian, maybe the better place to ask is on Italian wikipedia, ideally on a page that will attract readers who share your interest. That way, you'll most likely get a mother-tongue speaker which is what you need. How about here? (There might be even better places if you take half an hour to click round Italian wiki with a level of background knowledge which, I suspect, I cannot match!)

Success. Be well Charles01 (talk) 17:19, 27 March 2021 (UTC)