*tl;dr: We'll be stripping all content contained inside brackets from the
first sentence of articles in the Wikipedia app.*
The Mobile Apps Team is focussed on making the app a beautiful and engaging
reader experience, and trying to support use cases like wanting to look
something up quickly to find what it is. Unfortunately, there are several
aspects of Wikipedia at present that are actively detrimental to that goal.
One example of this are the lead sentences.
As mentioned in the other thread on this matter
lead sentences are poorly formatted and contain information that is
detrimental to quickly looking up a topic. The team did a quick audit
the information available inside brackets in the first sentences, and
typically it is pronunciation information which is probably better placed
in the infobox rather than breaking up the first sentence. The other
problem is that this information was typically inserted and previewed on a
platform where space is not at a premium, and that calculation is different
on mobile devices.
In order to better serve the quick lookup use case, the team has reached
the decision to strip anything inside brackets in the first sentence of
articles in the Wikipedia app.
Stripping content is not a decision to be made lightly. People took the
time to write it, and that should be respected. We realise this is
controversial. That said, it's the opinion of the team that the problem is
pretty clear: this content is not optimised for users quickly looking
things up on mobile devices at all, and will take a long time to solve
through alternative means. A quicker solution is required.
The screenshots below are mockups of the before and after of the change.
These are not final, I just put them together quickly to illustrate what
I'm talking about.
- Before: http://i.imgur.com/VwKerbv.jpg
- After: http://i.imgur.com/2A5PLmy.jpg
If you have any questions, let me know.
Associate Product Manager, Mobile Apps
Both mobile apps and web are using CirrusSearch's morelike: feature which
is showing some performance issues on our end. We would like to make a
performance optimization to it, but before we would prefer to run an A/B
test to see if the results are still "about as good" as they are currently.
The optimization is basically: Currently more like this takes the entire
article into account, we would like to change this to take only the opening
text of an article into account. This should reduce the amount of work we
have to do on the backend saving both server load and latency the user sees
running the query.
This can be triggered by adding these two query parameters to the search
api request that is being performed:
The API will give a warning that these parameters do not exist, but they
are safe to ignore. Would any of you be willing to run this test? We would
basically want to look at user perceived latency along with click through
rates for the current default setup along with the restricted setup using
User:Casliber has started an editing contest on the English Wikipedia
focused on improving or adding lead (intro) sections:
https://en.wikipedia.org/wiki/Wikipedia:Take_the_lead! (open until
this Sunday, January 31)
Notably, the invitation cites mobile readers as a motivation:
"Many articles on the English Wikipedia have deficient or poorly
written leads that do not summarise the article or present the
information in an engaging manner. With increased mobile phone usage,
this is becoming more of an issue, because mobile interfaces often
show the lead alone, with other sections collapsed."
I added the subsequent sentence, to support this point with data from
the analysis that JonR and I did some weeks ago for the
MobileWebSectionUsage data. This also lead to some interesting
community discussion in several places, and someone added that chart
. It's a great example of how we can support editors' work with
readership data beyond mere pageviews (I also submitted a Wikimania
talk about this topic earlier this month).
IRC (Freenode): HaeB
We've added Mediawiki parser content analysis to the content analysis
report that the Reading web team performed last quarter.
We also added the option to see the Gzip (lvl6) version of the report to
have a look at more realistic numbers (since traffic is gzipped in prod)
(see select box at the top).
No surprises, the results are pretty similar to the restbase analysis, in
that navboxes are around 14% of the content and references are around 50%.
*Request*: If you know about useless html markup emitted by the mediawiki
parser and would like to see what % of the content it accounts for, please
answer here or in the task with examples and we'll add it to the report
(like we did with restbase and the *extraneous markup*).
*Related phab task: https://phabricator.wikimedia.org/T123325
reply-all is hard...
---------- Forwarded message ----------
From: Erik Bernhardson <ebernhardson(a)wikimedia.org>
Date: Wed, Jan 20, 2016 at 12:14 PM
Subject: Re: [WikimediaMobile] Similar articles feature performance in
CirrusSearch for apps and mobile web
To: Joaquin Oltra Hernandez <jhernandez(a)wikimedia.org>
On Wed, Jan 20, 2016 at 7:45 AM, Joaquin Oltra Hernandez <
> I'd be up to it if we manage to cram it up in a following sprint and it is
> worth it.
> We could run a controlled test against production with a long batch of
> articles and check median/percentiles response time with repeated runs and
> highlight the different results for human inspection regarding quality.
> I can work this up i think. David and I have done some basic checks of a
few dozen articles and on average latency was around half using only
opening text. I'll work up something a bit more complete with a few
thousand articles across. One difficulty of this is that morelike
performance changes depending on cluster load. During the busy part of our
day morelike takes 50% longer than at the low points
It's been noted previously that the results are far from ideal (which they
> are because it is just *morelike*), and I think it would be a great idea
> to change the endpoint to a specific one that is smarter and has some cache
> (we could do much more to get relevant results besides text similarity,
> take into account links, or *see also* links if there are, etc...).
We've talked about a dedicated endpoint internally but haven't gotten
anywhere on it. I can fairly easily put together a cirrus specific api
endpoint, it's been hung up deciding if we should instead be putting the
api into core and building up some sort of abstraction around it. Putting
it into core would probably make more sense if we are doing more than the
basic morelike query.
> As a note, in mobile web the related articles extension allows editors to
> specify articles to show in the section, which would avoid queries to
> cirrussearch if it was more used (once rolled into stable I guess).
> I remember that the performance related task was closed as resolved (
> https://phabricator.wikimedia.org/T121254#1907192), should we reopen it
> or create a new one?
> I'll create a new one, some performance concerns were addressed there and
we did see a reduction in server work (average fetch latency cut in half).
Morelike is still accounting for around ~20% of server load, even though it
is only in the 700 qps range (vs 4k for fulltext and 8k for prefix. Note
these are after fanning out to shards, not the number sent to mediawiki).
I'm not sure if we ended up adding the smaxage parameter (I think we didn't
> should we? To me it seems a no-brainer that we should be caching this
> results in varnish since they don't need to be completely up to date for
> this use case.
I've been unsure about using smaxage on the search api, due to
fragmentation between how different clients use the api. After further
investigation I've perhaps been worried for no reason.
A relatively naive query in hive suggests in the span of 24h we could cut
morelike queries to the backend from 7.3M to 1.7M:
select sum(total), sum(deduplicated) from (select count(1) as total,
count(distinct requests.query) as deduplicated from
wmf_raw.cirrussearchrequestset where year=2016 and month=1 and day=10 and
requests.querytype = 'more_like' group by wikiid) x;
This tries to get a rough estimate on how that compares to the variance in
the way uri's are sent. I'm not sure how good of an approximation this is,
but the totals are similar enough it might be a good guess:
select sum(total), sum(deduplicated) from (select count(1) as total,
count(distinct uri_query) as deduplicated from wmf.webrequest where
year=2016 and month=1 and day=10 and uri_query LIKE '%search=morelike%'
group by uri_host) x;
In summary to resolve the current load issues we are seeing I will figure
out how to get these cached. I've created
https://phabricator.wikimedia.org/T124216 for discovery to figure that out.
Changing to opening_text would still provide a large benefit for latency on
non-cached pages. It may also help with relevancy, but that is hard to
guesstimate. I still think measuring click through rates could inform the
relevancy decision without too much programming overhead (depending on how
much work it is to add an AB test, i've been told its fairly painless in
On Tue, Jan 19, 2016 at 11:54 PM, Erik Bernhardson <
> ebernhardson(a)wikimedia.org> wrote:
>> Both mobile apps and web are using CirrusSearch's morelike: feature which
>> is showing some performance issues on our end. We would like to make a
>> performance optimization to it, but before we would prefer to run an A/B
>> test to see if the results are still "about as good" as they are currently.
>> The optimization is basically: Currently more like this takes the entire
>> article into account, we would like to change this to take only the opening
>> text of an article into account. This should reduce the amount of work we
>> have to do on the backend saving both server load and latency the user sees
>> running the query.
>> This can be triggered by adding these two query parameters to the search
>> api request that is being performed:
>> The API will give a warning that these parameters do not exist, but they
>> are safe to ignore. Would any of you be willing to run this test? We would
>> basically want to look at user perceived latency along with click through
>> rates for the current default setup along with the restricted setup using
>> only opening_text.
>> Erik B.
>> Mobile-l mailing list
FYI. Note this is captured this as something to think about in
https://phabricator.wikimedia.org/T123349. We're fully booked in Q3, but
this sort of thing seems worth examining for collaboration opportunities
for our future-facing roadmap.
---------- Forwarded message ----------
From: Lucie Kaffee <lucie.kaffee(a)wikimedia.de>
Date: Wed, Jan 20, 2016 at 5:57 AM
Subject: [Wikitech-ambassadors] Looking for small Wikipedias to test a new
feature for them
As part of my Bachelor’s thesis I worked on an extension called
https://www.mediawiki.org/wiki/Extension:ArticlePlaceholder over the last
One of the biggest barriers for accessing the knowledge Wikipedia provides
There are many topics that are only covered in few, big Wikipedias. People
who don’t speak any of these languages don’t have access to all the
information available potentially vital to them.
The Article Placeholder extensions aims at smaller Wikipedias to support
them in increasing access to data available on Wikidata. Article
Placeholders are automatically generated content pages in Wikipedia or
other mediawiki projects displaying data from Wikidata. They are clearly
not actual articles but an overview of data on a topic which does not have
an article yet. The design of the page and its content is under the control
of the local community via Lua and templates but we will provide defaults
so smaller Wikipedias can work with them without having to worry about the
technical side of it.
I have a test setup on Labs with an example for Ada Lovelace
The reader can find these pages by searching for a topic and gets results
if there is an Item on Wikidata with the respective label and/or alias.
The reader would benefit a lot since even if there is no article on a topic
yet, they will still have basic information provided in their language. But
it also might increase the numbers of editors due to increased usefulness
of that Wikipedia.
We are now looking for the first Wikipedias to support the extension by
deploying it and giving their input. I am still developing the extension
and the first Wikipedias to try it will naturally have a larger say in how
If your Wikipedia would like to give it a try please let me know. We would
start it as a beta feature.
Working Student Software Development
Wikimedia Deutschland e.V. | Tempelhofer Ufer 23-24 | 10963 Berlin
Phone: +49 (0)30 219 158 26-0http://wikimedia.de
Imagine a world in which every single human being can freely share in
the sum of all knowledge.
That‘s our commitment.
Wikimedia Deutschland - Gesellschaft zur Förderung Freien Wissens e. V.
Eingetragen im Vereinsregister des Amtsgerichts Berlin-Charlottenburg
unter der Nummer 23855 B.
Als gemeinnützig anerkannt durch das Finanzamt für Körperschaften I Berlin,
Wikitech-ambassadors mailing list
I posed some basic funnel analysis and Referer analysis for the Related
Articles / Read More functionality:
Short version: it seems to have relatively higher engagement on the mobile
web beta channel, but relatively lower engagement on the desktop web beta
Analysis and discussion on the feature is ongoing, but I wanted to share
out this data anyway.