Research talk:Increasing article coverage/Tool
The feedback for the tool goes here.
Feedback from Jonathan
1.add a 'give us feedback' button. It's fine if it just links to a talkpage.
2. Change "Articles Similar To (optional)". To "Enter an article title below to find similar articles to translate."
3. Change "Recommend" button label to "Get recommendations". Make the button blue (progressive): http://tools.wmflabs.org/styleguide/desktop/section-2.html#section-2.1
4. Change "Seed article" to "article title".
5. Don't pre-populate the 'To' field with a particular target language. Leave it blank until the user makes their selection. Maybe make the default option "select language" in grey text.
6. Not sure it makes sense to present the user with a list of recommendations when they first visit the page. I found myself wondering what those articles were, given that I had yet to specify my target language (I don't know Catalan >_<). Perhaps wait until the user has specified their target language before generating the first set of recommendations.
7. Instead of using "Articles Recommended for Translation" as the page title, make it the title of the section of the page that contains articles recommended for translation. Place it directly above the article list. This makes it clearer to the user what that list contains.
- 7.1. if it's not too hard, you might want to also add a sub-title that says something like "These articles from [source wiki] can be translated into [target wiki]"
8. Put the actual page title in the upper left of the menu bar. Maybe call it something like "Article recommendations" or "Recommender".
9. Add a "refresh" button that generates 10 new recommendations based on the current criteria. Call it "Load more recommendations" or "View more recommendations". Not sure how easy/hard this is, but it will be really useful for testing: we want the users to be able to explore as freely as possible so that we can get a better sense of what they want/need out of the tool. Jmorgan (WMF) (talk) 23:49, 20 August 2015 (UTC)
Ignore titles with redirects
The first few examples I clicked on to try translating seemed unlikely that English would be missing them... for example, I selected Bomba atomica (atomic bomb) on Italian Wikipedia - which was recommended to me. Of course, there is an article about this on English Wikipedia, but en:Atomic bomb is actually a redirect to en:Nuclear weapon where several typologies (including the atomic bomb) are discussed in the one article. This difference in the way that different languages scope their merging and separating of similar topics is a perpetual issue in deciding which Wikidata item to connect things to, and can't be solved with a neat 1-to-1 solution. In the mean time however... would it be possible to add a check for whether the output of a machine-translation for the article title already exists as an article or redirect on the destination language? And if so, ignore it as a suggestion? Wittylama (talk) 14:19, 29 April 2016 (UTC)
Some of the tool's documentation text seems to have been written by someone who didn't pass 7th grade. Example:
- "GapFinder helps you discover articles that exist one language but are missing another."
- Or English is not their native language. Give them a break! ✒ Bennylin 03:55, 3 May 2016 (UTC)
- Or that we just missed it, or both, or all. Thanks for pointing it out, Thekohser. I filed a ticket (T137957) for us to fix this. Thanks, Bennylin for encouraging a more friendly space, too. --LZia (WMF) (talk) 11:46, 16 June 2016 (UTC)
Please fix zh-yue, zh-min-nan and zh-classical
zh-yue, zh-min-nan and zh-classical Wikipedia still have no timetable to be moved to yue, nan and lzh, but the tool is already using yue, nan and lzh, so the result does not display normally. --Monsolo15 (talk) 12:32, 2 May 2016 (UTC)
- Hi Monsolo15. Thanks for flagging this. We have filed a task to fix this. You can follow it at https://phabricator.wikimedia.org/T138649
Feedback from Benny
- The "Search" field was not visible. I missed it at my first visit, and left the site thinking it still needs to be worked out some more, until a friend pointed out that there's a search field, then I got back and had my "Aha!" moment
- With "Search", I can think of several potential, one is searching within a category (and all of its subcategories) what articles are missing in the target language (Some statistics would also help). I tried searching "Category:Indonesia", and the result is not what I expected
- Search in a specific list (for example, list of Vital articles) for missing articles in target language would be awesome (visualization-wise). I tried searching "Wikipedia:Vital articles", and the result is also unexpected
- Remember my last language pair, so if I refresh or revisit the page, I don't have to enter the language pair everytime. Either use cookie, or via url (like Google Translate)
- Or, a button to easily refresh recommendations, so I don't have to refresh the page each time.
All in all, from my first try, it seems visually appealing to see the images, and not just the [red links] titles of missing articles, and give more incentives for translators to use the ContentTranslation tool. ✒ Bennylin 04:25, 3 May 2016 (UTC)
Hi Bennylin. Thanks for the input. Here are some responses:
- We have made some changes to the search bar to make it more discoverable. I'd say it can still become more dominant. We can even test with removing random recommendations altogether so the search becomes even more prominent (that change can also put more control in the hands of the user, as opposed to showing the user popular but still random articles.)
- Yes. We are planning to move the API behind this service to production, and filtering by lists or categories is something we can do much more easily in the production environment. We should do that.
- I answered this above.
- Great point re remembering the last language pair. We made a task to get to it: https://phabricator.wikimedia.org/T135208
- Would a refresh choice be more desirable or a "show more" at the bottom of the page?
If I see well, the tool ranks articles based on number of views in the source language wikipedia. However, relevance can differ a lot between languages. Wouldn't it be possible to take into account the interests of people using the target language? E.g. to rank taking into account the number of views from computers/browsers set with the target language? (E.g. for translations en->hu not to rank based on all views of the en article, but its views from computers/browsers with hu as the language)? In this way the results would have more relevance for the people using the target language. --126.96.36.199 18:45, 18 May 2016 (UTC) --Antissimo (talk) 18:46, 18 May 2016 (UTC)
- Hi Antissimo. This is a good idea. We also have tried in the past to take into account the number of other languages in which the article already exists in. This is usually a strong signal, as if many languages have the article, probably the specific language it's missing in should have it, too. Going beyond just the pageview count in source is a goal. --LZia (WMF) (talk) 23:09, 11 July 2016 (UTC)
Gapfinder bug - I chose Hindi as source and English as target and found several articles displayed but when I click them they run into trouble with no page to display. Browser independent I presume but I use a recent Firefox. Shyamal (talk) 10:00, 1 March 2017 (UTC)
- I have the same bug. Probably, because of russian language of interface. How can I switch it to English? --Alex Blokha (talk) 12:02, 31 May 2017 (UTC)
- Thanks for reporting. The issue has been captured in https://phabricator.wikimedia.org/T193575 and it has been fixed. Bmansurov (WMF) (talk) 19:37, 2 May 2018 (UTC)
Adding a location filter
I think this tool will serve better if it has another filter by number of views from a country or a region if that is possible. That way if some one is looking for articles to create in a particular language that is missing then they could also find the relevant articles according to their region. That way it is a better match on articles to be created. This offcourse could be an additional filter not necessarily embedded in the main logic. --Chinmayisk (talk) 06:20, 3 September 2017 (UTC)
- Thanks for the suggestion. It's been captured in https://phabricator.wikimedia.org/T193577.
Please add an option to export the suggested articles
After running some queries, I notice that in some cases, the articles exist but are not linked across languages. Having an export option for these articles allows a lot of flexibility. For example, you could plug the results to another tool (say Petscan) and further edit the Wikidata items for the pages.
Please consider adding such an option to export the results.
- Thanks for the suggestion. It's been captured in https://phabricator.wikimedia.org/T193576. Bmansurov (WMF) (talk) 21:56, 1 May 2018 (UTC)
For some of the articles, the links are broken. For example, I tried using Gapfinder for Hindi -> English. A number of the top articles had malformed links in the Hindi wikipedia link (when I tried to follow it)
- Thanks for reporting. The issue has been captured in https://phabricator.wikimedia.org/T193575 and it has been fixed. Bmansurov (WMF) (talk) 21:53, 1 May 2018 (UTC)
Unfortunately, this tool does not provide the option to search in all languages at once, please add this option as soon as possible and inform me..Mohmad Abdul sahib 19:21, 22 September 2021 (UTC)