Talk:Cross-wiki Search Result Improvements

Jump to: navigation, search

About this board

By clicking "Add topic", you agree to our Terms of Use and agree to irrevocably release your text under the CC BY-SA 3.0 License and GFDL

Suggestion: Reduce the weight of the description page and add a reporting tool

1 (talkcontribs)

One of the drawbacks of simply searching the file description and title is that it can be very unreliable and can easily be used to vandalize. This means that searching for "whore", surfaces obvious vandalism like this , . The image of a famous woman being "shown in results" for whore, or a "molester" (, Carl_Sagan)

Proposed solutions:

  1. Reduce the weight of the file description page - it can often be misleading and just plain bad especially on non-English wikis.
  2. Add a reporting tool to make it easier for people to report / remove vandalism - this will also help get more eyes on commons, and potentially somewhat reduce their workload.
  3. Prioritize media used locally in multimedia results - images used locally might be more relevant to the project and to the search.
Reply to "Suggestion: Reduce the weight of the description page and add a reporting tool"

Suggestion: Use Wikidata to fetch the main image for multimedia results

8 (talkcontribs)

Problem: As a non-English reader, search may often return irrelevant and inappropriate results due to using the image caption / description or inappropriate page-image.

Background: Consider if someone searches for monkey , and a man shows up due to the label or description of the image. In certain contexts this may be insulting. Aside from that, in non-English wikis, a lot of the times the multimedia results will yield wrong images due to the fact that the image descriptions are not translated.

Proposed solution:

When the search matches a wikidata item / alias use the image property as the primary image.


(Shows these:,,



DTankersley (WMF) (talkcontribs)

Thanks for the suggestion, but using wikidata to search Commons for images matching a search query is a bit different than what we're trying to do - right now.

Sometimes the images that are returned could indeed be offensive but we don't censor things in regards to how an image is tagged. We just do the search and then display the top most relevant results.

On other hand, this presents an excellent opportunity to edit images that are mis-tagged or mis-labeled to avoid them from showing in search results where they really don't belong. :)

Using your sample URL, I think if I was on frwiki and I was trying to do a search for a fish and I saw some math diagrams, I'd try a different search term. According to fr.wiktionary, the more common term for 'fish' in French is 'poiscaille' - which does return one image of a fish.

Or, being redundant in your search query by using 'poisson' and 'fish' returns an image of a fish, a fossilized fish and a large rock structure in the shape of a fish: Still not optimal. Clicking on the multimedia link (on the search results page) does indeed show more images of fishes: but it needs to have the dual search query terms. Again, not optimal at this time.

We have a new project that is starting soon that will help with this in the long term. The new project will be for structured data in Commons and we will be updating the API for searches like this. (talkcontribs)

Thanks for the response.

Perhaps it might be prudent to not let perfect be the enemy of good by implementing something like This would be a short term solution until the structured metadata project comes along and eventually replaces it. Currently people seem to ignore inaccurate images because they aren't really visible. The search interface doesn't surface them except on mobile, and on

Once this becomes widely deployed, you're very likely to frequently receive a bunch of "bug reports" of inaccurate multimedia content showing up as every search will potentially show up some image, audio or video, whereas currently only text is shown. Labeling these images won't work because the search engine seems to emphasize the image filename, instead of its description.

The idea is not to only use wikidata, but simply choose 1 image from it in addition to the normal ones that are already displayed. (talkcontribs)

One alternative idea is simply to pull the pageimage of the matching page. For example, for the "poisson" search string above, the first article in the search results is an exact match, and its page image would be perfect to illustrate the fish. Currently it seems that the search engine relies on simply searching the file namespace or commons for the article title, and it yields worse results than the first image in the article itself:

Even articles with no images may yield some illustration picked up from wikidata, if the task above is solved. (talkcontribs)

This is pretty bad even on english wikipedia, searching for something as common as a "leg"ː


The article itself contains a pretty strange image too (without the caption it was hard to tell what it is)ː

But wikidata has more easily recognizable images of legsː


DTankersley (WMF) (talkcontribs)

I agree that there could be better images for 'leg' but unfortunately, without more context about 'leg' it's hard to get really good results back.

For instance - using 'human leg' works pretty well:

or using 'dog leg' which shows a dogleg (zig-zag) and two images of a dog: (talkcontribs)

Well, that's partly true. Although leg is a pretty simple term, it wouldn't be that surprising if it at least managed to show a "table leg". There are even more cases where it fails for common terms:

The thing to remember is that many non-native English speakers use these resources. It may be quite easy for a native speaker to try and narrow their search but this isn't always possible for someone with limited knowledge especially in cases where a wikipedia / wikimedia project doesn't have resources in their native language. Relevant images would make it far easier to ensure that search results are relevant even before clicking any of them. Consider the discussion in this forum:

That individual would likely recognise the image faster than the text in the article. In fact the article description might just confuse them. A picture is worth a thousand words after all.

Anyway, hopefully structured commons will come eventually.

DTankersley (WMF) (talkcontribs)

Thanks for the real-life examples - they're always helpful. :)

I chatted with the Search team about this topic this morning - to be sure there wasn't anything that I was missing. Creating a new method, right now, to search for content on Commons will be a bit of an exercise in futility. Once the Structured Data team ramps up and gets their new format of metadata established, the Search team will incorporate it into the widely used CirrusSearch API and any extra work we do now will be trashed.

The goal behind the sister project search results is to give our readers and editors more information about their search query - to enable discovery into the other projects that maybe they didn't know about.

I'm confident that adding in the new sister project search results will aid in that discovery for millions of users - even though a better method of utilizing our search APIs will be coming in the near future.

For the example about 'paterna' -- if the 'inga feuilleei' term was used instead, it would indeed show lovely images of the fruit the user was hoping to find. Maybe those images could be tagged with the term 'paterna' by some very kind contributors, to make it easier for all to find?

Reply to "Suggestion: Use Wikidata to fetch the main image for multimedia results"

Should we limit the amount of languages we search in?

CKoerner (WMF) (talkcontribs)
  1. i.e.: only use the top 50 languages to implement this in?
  2. Or, only use the languages that we are detecting queries in an other language than the wiki the user is on?
RonnieV (talkcontribs)

''within the same language'' is stated clearly in Cross-wiki Search Result Improvements#A New Goal. So start easily.

If no results are found in the current language, it could be considered in a next step to search other languages, starting with those the user seems to understand (for instance: other language wiki's he has contributed to). Top-languages may result in more results, but also require more capacity (larger) and not everyone is capable of reading the top 5-50 languages.

Another option is to stick to related languages, for instance Roman if requested from a Spanish or Itailan wiki or German languages if the request comes from a Danish of Norwegian site.

Jeblad (talkcontribs)

If the number of languages are not limited, then smaller languages (and projects) will be swamped by hits from larger, and that would be very unproductive.

The actual languages should be the ones the user knows, ie. the babel list of languages. I don't think it should be limited to the language of the source wiki.

EncycloPetey (talkcontribs)

My comment above applies here as well: It really depends on the nature of the question. If someone is looking for the meaning of the Latin word ''vicesimanus'', Wiktionary information will be of most use, and it may not matter which language Wiktionary the results come from, as the word may only appear in a few projects, and might be illustrated with a picture, with a list of translations into other languages, or at least with an explanation in another language besides Latin. Likewise if someone is looking up the pronunciation of a word, or its syllabification for the purposes of hyphenating it, or synonyms. All of these features of a word may be presented on any Wiktionary, and may be found independently of the project language.

Quiddity (talkcontribs)

For myself (and perhaps most logged-in users), I think it would be ideal to allow quick searching from a specified list of wikis. (either by default, or as a new supplementary button in the Special:Search page).

For myself, as a somewhat meta-oriented and monolingual editor, I often want to be able to search (all of, or specific namespaces in):

  • English Wikipedia
  • Metawiki
  • Mediawikiwiki
  • Outreachwiki
  • WikimaniaXXXX

for example when I'm looking for documentation, or an essay, or an old presentation, or an old discussion.

Speculation: For multilingual logged-in users, I wonder if we could somehow re-use the info that Compact Language Links stores about languages we've purposefully visited? Or use a new shared storage location with a user-configurable override? (i.e. so that I can manually add a list of n languages.)

DTankersley (WMF) (talkcontribs)

Yes, we've discussed this in the past and we have a few tickets in Phabricator to investigate further, at least on the portal page. One ticket would add in a button or a link to switch the query string to search in different languages. Another ticket would display regional language links, but that probably wouldn't fit your need here. Just as an FYI, on the mobile apps, there is a way to preset languages to use for searching and it's fairly easy to switch between multiple languages.

However, you're requesting searching by project or namespace all at once, based on a pre-determined set of sites. I've added that idea into a new ticket to investigate how we could do this. I can envision a variety of ways we could do this, but we'd need to test and see which is more effective and intuitive to users (logged in or not).

Thanks for the suggestion!

Reply to "Should we limit the amount of languages we search in?"
Pginer-WMF (talkcontribs)

I tried the searching ogg files example, and some results are shown in a "multimedia card". Since all results are audio files all look the same (a speaker icon), I have to hover to get some more details based on the file name and make several clicks to actually listen to it.

Maybe we should consider a more informative way of preview results when they are audio files. This may involve including part of the name of the file, showing a relevant related image in addition to the speaker, playing the audio on clic/hover, or something else.

DTankersley (WMF) (talkcontribs)

Thanks for the suggestion - I've created a Phabricator ticket for this work.

Reply to "Preview of audio files"
Eduardogobi (talkcontribs)

The Plan section looks outdated, showing items (that should be done by now) in the future tense. Can someone from the Engineering or Discovery teams update it?

DTankersley (WMF) (talkcontribs)

Will do, thanks for the reminder! :)

Reply to "“The Plan” needs update"

Do we want these new search results to work across all Wikimedia projects?

CKoerner (WMF) (talkcontribs)
  1. For example, if I'm on Wikiquote, do I want to also see relevant search results from Wikivoyage, Wikipedia or Wikinews?
  2. Or, if I'm on Wikipedia, just show me results from other projects?
Jeblad (talkcontribs)

I wonder if specific projects have a given relevance for other projects, like Wikitionary have a higher relevance for Wikipedia, and a lower for Wikispecies. It will probably also change given the categorization of pages within the projects. Wikispecies has a high relevance for articles in Wikipedia within biography, but would have a low relevance for art.

If you do a search in a project, then the categories could be used as an indicator for how relevant (likely) some other project would be, given this specific result set. If a project is highly relevant, then the number of hits could be increased from 1 to 3 (just an example, use whatever number).

EncycloPetey (talkcontribs)

It really depends on the nature of the question. If someone is looking for the meaning of the Latin word ''vicesimanus'', Wiktionary information will be of most use, and it may not matter which language Wiktionary the results come from, as the word may only appear in a few projects, and might be illustrated with a picture, with a list of translations into other languages, or at least with an explanation in another language besides Latin. Likewise if someone is looking up the pronunciation of a word, or its syllabification for the purposes of hyphenating it, or synonyms. All of these features of a word may be presented on any Wiktionary, and may be found independently of the project language.

TJones (WMF) (talkcontribs)

I don’t think the average user searching English Wiktionary would be happy with a definition of a Latin term that was in Finnish, Russian, or Chinese—generally in any non-Indo-European language or any language that doesn't use the Latin alphabet. The lack of readable cognates makes those pages useless. Look at the Russian page for gato (Spanish "cat"). If you don't at least know some Cyrillic, you can't get much out of that page. Finnish gato is actually better than I expected, but only because there are some cognates (Espanja, Portugali, and substantiivi). You can translate those pages using your browser or online tools, but I think that's getting into the realm of “power users” unfortunately.

My intuition is that what most people want is results in the language of the project they are on, or projects in the same language. (Exception: when their query is clearly in another language. Exception to the exception: when they are on Wiktionary—which is where I often go for words I don’t know even when they are not in English.) Users could also use results in other languages they can read (which they need to specify or we need to surmise, say, based on browser settings). Only power users and researchers are going to dig into results for languages they don't know. This may change over time as machine translation gets better and people become more sophisticated about handling text in other languages—but I think most people aren't there yet.

I’m open to other opinions on user preferences and the typicality of any given use cases, of course!

However, there may be some technical limitations. We can’t index English Wiktionary both with all the other English projects and with all the other Wiktionary projects. Searching across all Wiktionaries without a shared index is probably too resource intensive to be practical.

EncycloPetey (talkcontribs)

Re: "Only power users and researchers are going to dig into results for languages they don't know." I disagree. During the time I was seriously active on Wiktionary, requests for translations into languages the user did not know were very common. We had daily requests for assistance.

TJones (WMF) (talkcontribs)

Interesting! Requests for translations into, say, Russian, seems very different from using a Wiktionary page in Russian (without machine translation).

Reply to "Do we want these new search results to work across all Wikimedia projects?"

How can we help users better understand the functionality of other wikis through verb-driven language?

MKramer (WMF) (talkcontribs)

Suggestion for language on boxes: of TIs there a way to indicate to users what they might find in each Wiki through a verb-driven phrase that prompts them to take an action? This could be hover-text on added tags or banners, or specific verb-driven language. For example, instead of "free dictionary" we may want to say "Look up the meaning of a word" which drives the user to take a specific actioin on Wikidictionary and also tells them what Wikidictionary does, instead of what it is. We might be able to test this, which would indicate might help users understand what these other resources are. srousers what h

DTankersley (WMF) (talkcontribs)

Yup - those all sound like great suggestions!

We've got several initial design mocks on the /Design page and these suggestions sound like a new mock candidate(s).

MKramer (WMF) (talkcontribs)

Thanks @DTankersley (WMF)! I'm not sure why additional letters were added to that question! Apologies for that!

I'm happy to write up verb-driven action sentences for each platform, although I don't know if we have standards floating around somewhere.

DTankersley (WMF) (talkcontribs)

Hi! A few examples would be cool to put on a mock, but I don't know if there are standards written up anywhere. :)


MKramer (WMF) (talkcontribs)

Awesome. Here are some examples, but these are drafts (and I am very open to feedback and edits...)

Wikidictionary - Look up the meaning of a word

Wikiquote - Discover quotes from thousands of people

Wikibooks - Read free textbooks about a variety of subjects

Wikipedia - Read a free encyclopedia that anyone can edit

Happy to contribute more if these are useful!

Reply to "How can we help users better understand the functionality of other wikis through verb-driven language?"

Will the display of the additional search results from other wikis encourage contributions from editors?

CKoerner (WMF) (talkcontribs)
  1. i.e.: if you search for Piazza del Duomo and don't see a Wikivoyage article about it (while I'm searching on Wikiquote), would that encourage you to start an article for it?
Jeblad (talkcontribs)

I believe so, but this is a research issue. (The problem is formulated as a question about convictions and feelings.)

EncycloPetey (talkcontribs)

If the intention is to replace the existing search function with the proposed new one, then the answer to the question will vary by project. If someone is searching from a Wikipedia, it may encourage work on the other projects. If someone is searching on Wiktionary, it may promote contributions to Wiktionaries in other languages. But a search from projects like Wikisource would pull contributors away from a project that is already struggling to gain new contributors, and would be irritating to users who are trying to search Wikisource itself.

DTankersley (WMF) (talkcontribs)

Hi @EncycloPetey - thanks for the feedback.

We're initially thinking of only putting the additional relevant search results that are gathered from the sister projects on the Wikipedia project itself. We'd like to see how this new feature is used by the broader community before determining if the new search results would work on the other projects (i.e.: searching Wikisource might not need or want results from Wikipedia or Wikivoyage).

Reply to "Will the display of the additional search results from other wikis encourage contributions from editors?"

How do we want these new, additional, relevant search results to be displayed?

CKoerner (WMF) (talkcontribs)

Some starter questions:

  1. Should the results from whatever wiki you're on to be shown first and then have an option to show more from other wikis?
  2. Should the additional results be inter-mixed with the local wiki results?
  3. Should the additional results be displayed off to the side (or maybe the bottom) of the results page?
  4. Should we have the option to turn off these other relevant search results (a user and/or project opt-out)?
    1. This could be a keyword search term or maybe a button for a visitor to click
    2. This could also be similar to the local: keyword that will only search for images on the local wiki and not Commons files, for instance.
  5. Would the additional results be best displayed as a list or a grid design?
  6. Should we include relevant metadata (images and/or a short description) with the search results?
  7. Do the results need to have the size of the article (i.e.: 848 bytes (104 words)) and the date it was created/modified?
  8. Should we indicate that clicking on a result will take you to another wiki project?
  9. How many results from other wikis should we show - 1, 2, 3, or more?
  10. Should we limit the existing method of displaying results from the wiki that you searched on?
    1. We currently show up to 10,000 results in a paginated manner, but testing shows that generally only the first 3 results are ever acted upon.
DarkShadowTNT (talkcontribs)

In my opinion:

Numbers 1 - 3, 5 & 8: Show the search results of the Wiki you're currently on, and then a seperate column besides it for results from other Wikiprojects (in a list), entitled like Results from other Wikiprojects or something like that. This means that the results of other Wikis can be found, instead of being somewhere at the bottom where most people don't look. Don't mix the additional results with local Wiki results, since it could look (in my opinion) very messy when that is the case. For example: English results mixed with languages that don't use Latin letters (like Arabic) and vice versa.

Number 4: It would be useful to provide an option to turn it off, in either the Preferences page or when viewing the search results (or maybe both?). In my opinion it is better to avoid using prefixes, especially when someone doesn't know about that.

Number 6: Images (the first image that is used on the page) would be useful, so that the (possible) reader globally knows what the subject of the page is. Kind of what currently happens when searching on the mobile Wikipedia site.

Number 9: I think showing the most relevant 1 to 3 results at the top, and then at the bottom of those results the option Show more results (hot-loaded (not loaded when loading the search page for the first time (or something like that)), so that searching stays as good as fast as it was before the implementation.

Number 10: I don't even know why you would look through 10,000 pages to find a subject... I think capping it at about 500 to 1,000 should be more than enough, since the testing shows that generally only the first 3 links are being used.

BTW, this message may contain some Dunglish

DTankersley (WMF) (talkcontribs)

Thanks for the feedback!

I think we can add in a setting in the preferences that could be used to turn on (or off) the cross-wiki search results for the more experienced users. But I do like the idea of having it visible on the page to turn it on/off easily.

Jeblad (talkcontribs)

Show the best match above a threshold for each of the other projects, in a single list, and with a "more" link which opens additional entries inline.

An alternative is a "more like this" and then opening additional entries on the side. Think of it as a sideways stack, where l-to-r languages have the new column on the right. That would make it possible to drill down in a relevance ranked set, simply by clicking in the additional columns.

If you select (click) on a page from an other project, then that project should take precedence in the new column, but when you have selected two or more projects then all of them should take precedence.

It follows from this that the page from column A could be a Wikipedia article and the page from column B could be a page from Wikitionary. The resulting relevance ranked column should then be pages that rank according to the selected pages from those two projects.

DTankersley (WMF) (talkcontribs)

Hi @Jeblad - that sounds quite intriguing!

Our initial focus is to gather that first selection of related (or "more like this") articles from the sister projects to display the first time a user enters a query into the search box.

Your idea is to expand this process to display a second set of related results on the page result (that the user selected) and expanding to include results from the first wiki and the second clicked-through to wiki site. Sounds cool!

We'll keep your idea on the backlog - as once we've launched the first iteration of showing additional relevant articles across wikis, we'll need to closely monitor it to see if our community likes it. :)

Jeblad (talkcontribs)

I don't think a user has a clear idea of where a "more like this" starts, and thus the first level (s)he open are probably pretty close to a simple "more" in the users mental model. That could imply that it is wise to start with a simple solution, and then later on twist the idea into full relevance ranking.

Forgot to mention that on smaller screens the additional columns can be rendered and then slide in horizontally, and the old columns sliding out, thereby signaling to the user how (s)he can navigate back to previous results. On large screens (4k++) the columns can stack up side by side.

Reply to "How do we want these new, additional, relevant search results to be displayed?"

Which wiki projects would be included in the search results?

DTankersley (WMF) (talkcontribs)

Currently, the Discovery Search team is focusing the new cross-wiki search results to be gathered only on the sister wiki projects.

For example, when searching on fr.wikipedia - cross-wiki results could be shown from fr.wikivoyage, fr.wikiquote, fr.wikiversity, fr.wikisource, fr.wikinews, fr.wikispecies, fr.wiktionary and fr.wikibooks - if results are found in those sister projects.

Trizek (WMF) (talkcontribs)

Wikidata can show really pertinent results.

Also, have you noticed Extension:ArticlePlaceholder?

DTankersley (WMF) (talkcontribs)

Hi @Trizek (WMF) - I chatted with the team and we feel that Wikidata is very excellent and a lovely pivot that we can maybe use for searching across languages- it might be a good candidate as a companion to our language detection efforts.

But for the cross-wiki search functionality, we really don't have any control on what data is actually searched for - because everything is merged into one big field. So, we don't think Wikidata would be a good fit for this particular functionality.

Jeblad (talkcontribs)

Note that some sister projects are very small and using them will give very poor quality results. With few and poor quality pages on a sister project those will end up on a lot of searches on some other project. That could trigger a lot of discussions, and prompt inclusion of a user option to turn the results off.

A solution could be to set a threshold on some kind of quality on the included hits, thereby limiting the number of pages where a certain hit can show up.

This could also give more incentive to clean up the low quality pages on other projects, so it could be "a wanted effect" even if it will create a lot of discussions.

Reply to "Which wiki projects would be included in the search results?"