I didn't make my first edit until April 2009. That was for the nascent Ada Lovelace Day.

Since then, I've racked up a bit over 600 edits which simultaneously feels like a lot and barely anything.

Every edit gives you a crude representation of how many characters you've deleted or added. If I've done my sums right, I've added about 86k letters to Wikipedia and deleted about 25k. So a net addition of 61K characters.

That feels like a worthwhile contribution to the commons.

She was the youngest and first woman to receive a British House of Commons Shield.

That sounds like a fantastic achievement! But, and I hate to bring fact-checking to an argument, what is it?

Looking around the web, I could only find three mentions of this no-doubt prestigious award.

1. A newspaper clipping of a puff-piece which made the same claim about the same person.
2. Reference to shields which are erected in memory to MPs who were killed in service.
3. A religious leader who claimed to have received the same award.

If this is a well-regarded way to recognise and honour people, shouldn't there be more about it online? But, perhaps, not everything to do with our democracy is available on the web. To the FOI machine, Robin!

I wrote to the House of Commons to see if they could shed some light on the matter.

As ever, their Freedom of Information team wrote back quickly:

In the first instance, please note that we are not aware of a “House of Commons Shield” award, and this is not something supplied by or handled the House of Commons Administration. However, from publicly available resources our understanding is that this was an award which has been supplied in the past, and we believe this may have been handled by individual MPs themselves as opposed to the House Administration. We have therefore checked with the following teams within the House Administration to see if they held any information within the scope of your request:

• The Speaker’s Office, who hold details of official awards provided by the House Administration, as well as those nominated for formal Honours.
• It is possible that these awards were provided by at a sponsored event organised by MPs, so we have checked with the House’s Events’ team on this.
• It is also possible that these awards were provided at room bookings made by MPs, so we have checked with the House’s Access & Services team on this also.

However, these teams have confirmed that they do not hold any information regarding this award. We therefore hold no information within the scope of your request.

They suggested that I contact any MPs who might have handed out the award. There was only one (former) MP I could find so I dropped them a polite email asking WTAF was going on.

Please can help me solve a small mystery? I'm trying to find out more information about an award called "The House of Commons Shield".

I understand that you presented the award to [redacted] in 2009. I'd like to understand a little more about what the award is.

The House of Commons have told me that this isn't an official award as such and they suggested I contact MPs who may have presented it.

A few people claim to have been given this award but you are the only named person I can find who has presented it.

Is "The House of Commons Shield" a specific award that MPs can present? Is it a recognised term? Are there any criteria used when judging who should get it? Or is it a piece of memorabilia which has been mistakenly claimed as an official trophy?

Any assistance in figuring this out would be gratefully received.

Sadly - but predictably - they elected not to enter into correspondence with me.

If you have any information about what the "House of Commons Shield" award is - please drop a note in the comments.

At the time of writing, the Wikipedia entry for "frequency illusion" said:

The name "Baader-Meinhof phenomenon" was coined in 1994 by an online message board user, who, after mentioning the name of the German terrorist group Baader-Meinhof once, kept noticing it, and posted on the forum about their experience. This led to other readers of the message board sharing their own experiences of the phenomenon, leading it to gain recognition. It was not until 2005, when Stanford linguistics professor Arnold Zwicky wrote about this effect on his blog, that the name "frequency illusion" was coined.

But, crucially, there were no citations to support this supposed origins.

A quick footle round the Internet led me to the Dictionary.com entry which said:

The name of the phenomenon is thought to have been coined by Terry Mullen, who explained his experience with it in a 1994 letter to a Minnesota newspaper.

Interestingly the date is the same, but it is now credited to a letter to a newspaper rather than a forum or message board.

A 2005 blog post about the Baader-Meinhof phenomenon contains a comment which purports to be from Mullens.

I coined the term “Baader-Meinhof Phenomenon” 20 years ago, or so. [...] I alerted the ‘public’ to the Baader-Meinhof Phenomenon through a series of letters I submitted to the “Bulletin Board” in the St. Paul Pioneer Press. I submitted them under the pen name “Gigetto on Lincoln,” a reference to my wheaten terrier named Einstein, nicknamed Gigetto.

We can see now where Wikipedia gets its assertion that it was an "online message board" as they were frequently called "Bulletin Boards" - but this appears to have been a newspaper-based board.

In 2019, there was a short online discussion talking about the origin of the phrase.

The Dread Pirate Bob

@RandomlyBob

Allow me to introduce you to Dan Kelly, formerly the St. Paul Pioneer Press feature editor who published the original reader submission naming the Baader-Meinhoff Phenomenon. He tweets as @BBonward.

Ben Zimmer

@bgzimmer

Now on @LanguageLog: "Frequency illusion," a coinage of Arnold Zwicky, has made it into the @OED: "a quirk of perception whereby a phenomenon to which one is newly alert suddenly seems ubiquitous." languagelog.ldc.upenn.edu/nll/?p=42180

---

---

Ben Zimmer

@bgzimmer

@RandomlyBob @BBonward Very cool! But will we ever know the real name of the reader known as "Gigetto on Lincoln," who wrote that submission? Lexicographical glory awaits...

---

That led me to the St. Paul Pioneer Press new Bulletin Board.

From March of 1990 until October of 2016, Bulletin Board ran every day in the St. Paul Pioneer Press. Now it is an independent production, running online, here at BBonward.com — and on Sundays in the Pioneer Press (twincities.com).

Aha! With that information, I found the original article from 1994!

So I corrected Wikipedia.

Now, does anyone have a scan of the original paper?

At first glance, it appears to be a normal London Tube map. But look closer...

Cool! But there is something about it which has always bothered me. Each Tube line represents a theme - therefore, a station at the intersection of multiple lines should be represented by someone who matches all of those themes.

For example, here's Baron's Court - the intersection of the Explorer line and the Saint line - represented by Saint Ursula. She is just an saint - she has nothing to do with exploring. This artwork is wrong!

So, can we write something to query Wikidata to generate a more accurate artwork?
Because accuracy is my aesthetic.

Jump straight to the finished product and skip all the geeky details!🔗

A brief guide to SPARQL

Wikipedia holds structured data about people and things. It uses SPARQL to query that data. It is a bit complex to use, but a valuable skill.

For example, this query finds people who are explorers and also saints.

SELECT DISTINCT ?person ?personLabel WHERE {
  ?person wdt:P106 wd:Q11900058 . # People whose occupation (P106) is explorer (Q11...)
  ?person wdt:P411 wd:Q43115      # People whose canonization status (P411) is sainthood (Q43...)

  SERVICE wikibase:label {
    bd:serviceParam wikibase:language "[AUTO_LANGUAGE],en" .
  }
}

The result is just four people. And that's where the problem starts. Simon Patterson chose categories for the lines which often don't have any intersections. There is no one who is an Italian Artist who is also a Saint and also a Footballer.

So, to create a more accurate artwork, we'll need different categories.

Nodes

One of the first things we need to do is understand the Tube map as a graph - with stations as nodes and lines as edges.

We want to know:

1. Which stations are on which lines
2. Which stations are on multiple lines
3. Which station has the most lines
4. How many stations are on each line

Thankfully Mark Dunne has done lots the hard work for us, and provided a great tutorial. Sadly, the data are about 5 years out of date.

Alternatively, the TfL API has lots of the information we need. Here's the call for all the stations on the Bakerloo line - https://api.tfl.gov.uk/line/bakerloo/stoppoints

Let's throw some Python down to grab the data we need. First, how many stations are there on the Bakerloo line?

import requests
r = requests.get("https://api.tfl.gov.uk/line/bakerloo/stoppoints")
stations = r.json()
count = len(stations)
print("There are " + str(count) + " stations on the Bakerloo Line")

Next, let's get the lines for each station:

import requests
r = requests.get("https://api.tfl.gov.uk/line/bakerloo/stoppoints")
stations = r.json()

for station in stations:
    stationName = station["commonName"]
    lineGroups   = station["lineModeGroups"]
    for lineGroup in lineGroups:
        modeName = lineGroup["modeName"]
        if (modeName=="tube"):
            lineCount = len(lineGroup["lineIdentifier"])

    print(stationName + "," + str(lineCount))

The line names can be found at https://api.tfl.gov.uk/Line/Mode/tube

Brief survey of the problem...

270 Tube Stations(!) across 11 lines. King's Cross St Pancras has the most lines - 6.

There a few anomalies in the data. It lists Edgware Road as two separate stations - even though it's really one station. The same problem is present on Hammersmith and Paddington. Cleaning data is "fun"...

The categories are also challenging. This is how many times the Bakerloo line intersects with the other lines

'bakerloo': {
    'circle': 3,
    'hammersmith-city': 1,
    'jubilee': 2,
    'metropolitan': 1,
    'northern': 4,
    'district': 2,
    'central': 1,
    'victoria': 1,
    'piccadilly': 1,
    'waterloo-city': 1
},

That is - the Bakerloo line touches every other line at least once. As do the Northern, Central, and Jubilee lines. Those lines will need to contain some very broad categories.

Back to Wikidata

So, we want to replace each station's name with a human's name. We need attributes which are wide-spread enough to get good coverage in the data - and quirky enough to be interesting. I'd also like to keep some of the original categories:

I suspect there's a way to interrogate SPARQL to find a list of categories based on a graph - but I'm not clever enough to do that. I started off with an entirely arbitrary set of attributes:

• Academy Award Winners
• Left-handed people
• Nobel Prize Winners
• People born in London
• Educated at UEA (the university where my wife and I first met)
• Female Computer Scientists
• Saints
• Explorers
• Journalists
• Sinologues
• Comedians

Here's the query for Comedians who were educated at UEA and were born in London:

SELECT DISTINCT ?person ?personLabel WHERE {
  ?person wdt:P69 wd:Q1045828 .
  ?person wdt:P106 wd:Q245068 .
  ?person wdt:P19 wd:Q84
  SERVICE wikibase:label {
    bd:serviceParam wikibase:language "[AUTO_LANGUAGE],en" .
  }
}

One result - Doc Brown. There are no saints who have won an Oscar, and data about left-handed people is suspiciously absent. The categories will have to be completely rejigged.

Keep It Simple, Stupid

I figured the easiest thing to do would be to start from a well data'd individual and work backwards from there.

SELECT DISTINCT ?person ?personLabel WHERE {  
  ?person wdt:P108 wd:Q35794  . #Employed by Cambridge University #Bakerloo
  ?person wdt:P19  wd:Q84     . #Born in London                   #Circle
  ?person wdt:P463 wd:Q123885 . #Member of the Royal Society      #Hammersmith&City
  ?person wdt:P106 wd:Q121594 . #Professor                        #Waterloo&City
  ?person wdt:P106 wd:Q205375 . #Inventor                         #Metropolitan
  ?person wdt:P106 wd:Q81096  . #Engineer                         #District
  ?person wdt:P106 wd:Q4964182. #Philosopher                      #Piccadilly
  ?person wdt:P106 wd:Q11063  . #Astronomer                       #Victoria
  ?person wdt:P106 wd:Q170790 . #Mathematician                    #Jubilee
  ?person wdt:P106 wd:Q82594  . #Computer Science                 #Northern
  ?person wdt:P106 wd:Q188094 . #Economist                        #Central

  SERVICE wikibase:label {
    bd:serviceParam wikibase:language "[AUTO_LANGUAGE],en" .
  }
}

The result of that query is the inventor of steampunk, Charles Babbage!

We can do a reverse query. Given these people, which common properties do they have?

SELECT ?property ?propnameLabel ?value_Label
where
{ 
  wd:Q46633 ?property ?value . #Babbage
  wd:Q7259  ?property ?value . #Lovelace
  ?propname wikibase:directClaim ?property . # constrain to directClaims
  SERVICE wikibase:label { bd:serviceParam wikibase:language "[AUTO_LANGUAGE],en".
                         ?value rdfs:label ?value_Label .
                         ?propname rdfs:label ?propnameLabel .}
} order by ?property

(Thanks to TagishSimon for the help)

This is where things got trickier! Most of the major intersections didn't have any candidates other than Babbage - truly a Renaissance Man! - so I expanded "born in London" to "born in the UK".

We can't use ?person wdt:P27 wd:Q145 becuase that only covers the current United Kingdom - not The Kingdom of Great Britain (1707–1801) nor The United Kingdom of Great Britain and Ireland (1801 to 1927)

The correct query seems to be ensuring the place of birth is within the current administrative territory of UK:

?person wdt:P19 ?pob . ?pob wdt:P131* wd:Q145 .

But you can also use a UNION

{?person wdt:P27 wd:Q145} UNION {?person wdt:P27 wd:Q174193} UNION {?person wdt:P27 wd:Q161885} .

That got closer - but still not enough.

Unions

Wikidata is fickle. Someone may have an occupation as a "computer scientist" or they may work in the field of "computer science". SPARQL eschews the or operator, and uses UNION:

{?person wdt:P101 wd:Q21198} #Field of Work CS
UNION
{?person wdt:P106 wd:Q82594} #Occupation CS

I'm beginning to see why the original artist was more liberal in his accuracy!

Sorting

I want the map to contain notable people. There are a couple of ways to assess the "notability" of a Wikidata subject. I've chosen to use "sitelinks" - that shows how many languages their article is available in. It's a crude, but quick method.

Here's it in action:

SELECT DISTINCT ?person ?personLabel ?sitelinks WHERE {
  ?person wdt:P106 wd:Q205375 . #Inventor
  ?person wikibase:sitelinks ?sitelinks .

  SERVICE wikibase:label {
    bd:serviceParam wikibase:language "[AUTO_LANGUAGE],en" .
  }
} ORDER BY DESC (?sitelinks)

Correcting for Bias

Wikipedia has an acknowledged male bias. So I used SPARQL's FILTER property to great effect:

FILTER ( !EXISTS{ ?person wdt:P21 wd:Q6581097 }) 

It says to return anyone without the sex/gender of "Male". (Yes, I know things are a bit more complicated than that - but this is a good way to return women, intersex people, agender, transgender folk etc).

If no non-men were returned, I repeated the search but omitted the filter.

Because I used "Born in the UK" as a filter, there is probably a bias towards white people. And people who become Professors or members of the Royal Society may also be the product of a biased society. There are many other filters and categories I could have chosen - and I hope some of you will create maps for your own cultures and societies.

P-p-p-pickup Some Python

There are several Python libraries for SPARQL, I used sparqlwrapper.

Here's a sample query

from SPARQLWrapper import SPARQLWrapper, JSON

sparql = SPARQLWrapper("https://query.wikidata.org/sparql")
sparql.setQuery("""
    SELECT DISTINCT ?person ?personLabel ?sitelinks WHERE {
        ?person wdt:P106 wd:Q81096 . #Engineer
        ?person wdt:P20  wd:Q84    . #Died in London
        ?person wikibase:sitelinks ?sitelinks .
        SERVICE wikibase:label {
            bd:serviceParam wikibase:language "[AUTO_LANGUAGE],en" .
        }
    } ORDER BY DESC (?sitelinks)
""")
sparql.setReturnFormat(JSON)
results = sparql.query().convert()

for result in results["results"]["bindings"]:
    print('%s %s %s' % (result["person"]["value"], result["personLabel"]["value"], result["sitelinks"]["value"]))

Plotting onto an image

Let's leave who and what we select for now, and work out how we draw the eventual results.

We could do all sorts of clever things plotting out locations - but I decided to cheat!

There is a fully semantic SVG of the tube lines (Thanks to Oliver O’Brien for finding it).

I removed all the lines I didn't want, then I was able to search & replace station names with my preferred text.

Search And Replace

Well... sort of! Charing Cross Station has an ID of 940GZZLUCHX.

On the map it is:

<g id="s-940gzzluchx_label">
    <g id="s-940gzzluchx_label_1_">
        <text transform="matrix(1 0 0 1 515.8999 487.6963)">
            <tspan x="0" y="0">Charing</tspan>
            <tspan x="4.3" y="4.5">Cross</tspan>
        </text>
    </g>
    <polygon id="s-940gzzluchx_nr" fill="#EE3124" points="514.3,487.7 512.6,486.9 515,486.9 515,486.4 512.5,486.4 513.8,485.8 515,485.8 515,485.3 513.8,485.3 512.2,484.6 511,484.6 512.7,485.3 510.2,485.3 510.2,485.8 512.7,485.8 511.4,486.4 510.2,486.4 510.2,486.9 511.5,486.9 513.1,487.7 "/>
</g>

Whereas the SVG element for North Wembley is just:

<text id="s-940gzzlunwy_label_2_" transform="matrix(1 0 0 1 282.1489 289.1079)">North Wembley</text>

Again, I'm not quite clever enough to work out a way to reliably find the inner text for an element which may be inside (or not) several other similarly named elements.

So a lot of repetitive search-and-replace it is. *sigh*

I also need to manually place some of the station names, because they're a different length to the originals. *double-sigh*

Putting it all together

I have great pleasure in revealing to you "The Great(er) Bear"!

See the full sized version.

Copyright

OK gang, turns out that copyright law is even trickier than computer code! I've spoken to Simon Patterson and he is happy for me to host a not-for-profit version of this piece of art which is heavily indebted to his original.

TfL has been litigious in the past when it comes to derivative maps. I tried contacting them several times, but didn't receive any clear answers as to whether I could do this.

The data that I used to generate the art is "Powered by TfL Open Data" and provided under OGLv2. It may contain OS data © Crown copyright and database rights 2016.

The original font is tightly controlled. So I've used a freely available font called Hammersmith One which is broadly similar.

Lots of people create modified tube maps:

• The Literary Tube Map - putting authors on the map
• Tube Map of Beer Prices
• Decibel levels on the Tube
• House Price Tube Map
• And no doubt many more.

If you want to build your own version of my modified map, all the data are on my GitLab!

Details

Here are a few interesting close-ups of the map - they may be different from the final version.

Errata

• The data in Wikidata may be incorrect or incomplete.
• I originally didn't restrict it to just humans! So a few weird entries snuck in. Using ?person wdt:P31 wd:Q5 . corrected that. But I'm wondering if anyone on the map is fictional...!
• Due to timeouts and my crappy coding, I ran the code over several passes on different days. If you run the code, you might get different results.
• I didn't use people's names in their original language, I had to back-fill them. I probably missed some. I should have used P1559.
• Even after lots of jiggling of categories, one or two stations kept coming up blank. So I manually added in a few people. Can you spot who they are?
• Some people's names were too long for the allotted space, so I have swapped a few people around. Better code would try to keep name length as close to the original as possible.
• There's no (intentional) ordering. It might be nice to put people on the line in order of, say, year of birth.
• Similarly, there's almost no relation between the people and the places. Although I've contrived to put the author of Mary Poppins somewhere special!
• The Hammersmith One font only has a basic set of characters - so non-European languages (and some accents) are in the default font.
• The Elizabeth Line / CrossRail hasn't opened yet. I suspect it will be much harder to produce a new map once it goes live. Similarly the DLR and Overground lines are excluded.
• The SVG renders well in Firefox, and seems to work OK in Chrome. Please let me know of any glitches.
• I've also added a couple of Easter Eggs. Enjoy finding them!

Thanks

Mainly to my wife, Liz, for being very patient with me while I swore at my code.

I am indebted to Simon Patterson for his incredible and inspirational artwork. When it was created in 1992, Wikipedia did not exist. Linked Data stores were in their infancy. It would have been close to impossible to create a semantically correct map. Nothing in my version is intended to take away from Patterson's work and creativity.

I began to wonder - how often are Tweets used in citations?

It's possible to search for your own Tweets using this (somewhat obscure) link:

https://en.wikipedia.org/w/index.php?title=Special%3ALinkSearch&target=twitter.com%2Fedent

Just edit the end of it to see if you, or your friends, have been cited. Note - the username is case sensitive, so "Edent" isn't the same as "edent".

For example, we can see where Cory Doctorow is cited:

https://en.wikipedia.org/w/index.php?title=Special%3ALinkSearch&target=twitter.com%2Fdoctorow

Aha! The page on the New Zealand Internet Blackout references:

Cory Doctorow NONCONSENSUAL BLUE TICK

@doctorow

Success! I'm blacked out in solidarity with Kiwis whose net has been pwned by American entertainment giants: http://creativefreedom.org.nz/

---

Ok, so which Twitter user has been cited the most? TO THE API, ROBIN!

Wikipedia's own help pages are a little lacking, so I went to the help pages of the software which runs Wikipedia - MediaWiki.

We want to search for external URLs which point to Twitter and have a namespace of 0 (that means they're articles, not talk pages). We can grab a maximum of 500 results at a time, using JSON, and we want to include "www.twitter.com" and "twitter.com". Here's what we use.

https://en.wikipedia.org/w/api.php? action=query& list=exturlusage& eunamespace=0& eulimit=500& format=json& euquery=*.twitter.com

Run it yourself to see the results.

Limitations

• This only gets 500 results at a time - to paginate through, we add euoffset=
• This only searches the English Wikipedia.
• There are roughly 17,800 links to Twitter from English Wikipedia returned by the API.
• The majority of citations just point to a Twitter user's page - not to a specific Tweet.
• Some of the returned Tweets use the obsolete HashBang URls (e.g. http://twitter.com/#!/007/status/133679555167784960.)
• Some Tweets have been deleted.

Crappy Python!

import urllib2
import json
from collections import Counter

#euoffset=17800
api = "https://en.wikipedia.org/w/api.php?action=query&eunamespace=0&eulimit=500&format=json&list=exturlusage&euquery=*.twitter.com"

euoffset = 0

words = []

while euoffset < 17500:
    try:
        site_data = json.load(urllib2.urlopen(api + "&euoffset=" + str(euoffset)))
        #   Itterate through
        for element in site_data['query']['exturlusage']:
            #   Remove HashBangs #!
            #   Lowercase everything
            twitterURL = element['url'].replace("/#!","").lower()
            twitterUser = twitterURL.replace("http://twitter.com/","")
            twitterUser = twitterUser.replace("https://twitter.com/","")
            twitterUser = twitterUser.replace("@","")
            slash = twitterUser.find("/")
            if slash > 0:
                twitterUser = twitterUser[:slash]
            print twitterURL
            # print twitterUser
            words.append(twitterUser)
    except urllib2.URLError:
        print "Unable to retreive data"
        sys.exit()
    euoffset += 500

#   Most cited user
word_counts = Counter(words)
print word_counts

The Results...

The most cited Twitter users are...

1. LaLiga A Spanish Football competition - 105
2. Lea Michele An American singer / TV actor - 54
3. Guldbaggen The Swedish equivalent of the Oscars - 35
4. Kevin Tancharoen An American movie director - 12
5. PRESTO card Ottawa's public transit smartcard - 10
6. ICE T An American rapper - 10
7. Northern Pride RLFC A British rugby team - 10
8. 穴井勇輝（勇吹輝） A Japanese Actor - 9
9. NICKI MINAJ An American singer - 8
10. Counting Crows An American band - 8

And the most cited individual Tweet?

Linked to from lots of Lost Girl pages.

What Have We Learned Today?

Wikipedia does have a large amount of pop-culture (do we need hundreds of words on My Little Pony Characters?)

Twitter, unsurprisingly, has limited utility as an encyclopædic source - it's great for breaking news and ephemeral events, but it's fragile and lacks depth. There are very few occasions where Twitter would be the sole, and canonical, source of information [Citation needed].

Learn more on Andy Mabbett's blog or follow #WikiVIP on Twitter.

Get About A Minute as soon as each episode goes live. Stick this Podcast Feed into your podcatcher Or you can Subscribe on iTunes

Intro music "Gran Vals" performed by Brian Streckfus. Stopwatch Icon by Ilsur Aptukov from The Noun Project.

This podcast is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.

That said, I've not been able to devote as much time as I would like to - nor have other project members. We'd like to see it blossom and grow but, sadly, our resources are too limited to be able to make much impact.

So, it with great pride that we're happy to announce that QRpedia has been acquired by Wikimedia!

As much as I'd have loved this to have been a Silicon Valley exit with billions of dollars being splashed around - we agreed that the sensible thing to do would be to grant all the IP to Wikimedia and have them operate it in perpetuity.

Over the next few days we'll be transferring the website and all the other assets to the new owners - so there may be a little down time. But I'm confident that this will QRpedia to grow and grow.

The source code remains open for anyone who wants to see how it works and/or implement a similar system.

Derby Museum using multilingual QR codes from Andrew James Sykes on Vimeo.

In 2011 Wikipedians wrote and translated 1200 new articles to allow the museums objects to read in over a dozen languages using QRpedia codes.

You can see all the articles (and their QRpedia codes) - there is also a list of articles which need translating.

QRpedia's Name

There is some confusion about QRpedia's name.

Торчковій музонъ

@dslrave

Replying to @QRpedia@QRpedia please answer me, QRpedia = QR + wikipedia or QR + encyclopedia? I need it as a prooflink for [[ru:QRpedia]].

---

The answer is very simple. The "pedia" isn't from "Encyclopedia". It isn't from "Wikipedia". It's an acronym.

• P - Potentially
• E - Every
• D - Device
• I - Interlanguage
• A - Access

The "QR", of course, standing for "Quick Response".

I hope that clears up the matter ;-)

All these languages are valuable and hugely important to their communities. Many have a Wikipedia version written in their language.

Unfortunately, very few phones support these languages.

This poses a problem for QRpedia. They way the system works is this:

1. Read the phone's language
2. Look for a suitable translation in Wikipedia
3. Return the correct article
4. If a translation doesn't exist, return a list of available articles

Suppose The National Library of Wales has a QRpedia code for the Black Book of Carmarthen. A Welsh speaker will probably wish to go to the Welsh version of the article. However, their phone does not support the Welsh language (unless it is a Samsung S5600) and is set to English.

QRpedia, therefore, redirects them to the English version and doesn't give them a chance to read in their native language.

This is a problem we have faced with both Catalan and Norwegian.

Catalan

Catalan faces the very same problem as Welsh does in the previous theoretical example. Many people speak it but, because it's rare for a phone to support it, their phones are set to Spanish.

This was how we solved the problem:

• If the QRpedia code was for a Catalan page (ca.wikipedia)...
• If the phone's language is Catalan (CA) take them to the Catalan Wikipedia.
• If the phone's language is Spanish (ES) take them to a language choice screen - they can then select between Spanish, Catalan, or any other available language.
• If the phone's language is anything else (say EN) take them to the article in their language.

QRpedia doesn't store the user's language choice - so the user has to choose every time the scan which language they want.

The reasons we don't store the language choice is that it would be very hard to undo if the user made a mistake, or ever wanted to change their language.

Norwegian

The Norwegians have two languages - Bokmål and Nynorsk.

The standard language codes are NB and NN. However, most phones only support NB - with the language header of NB-NO. To complicate matters, the NB Wikiedia is located at NO.wikipedia!

So, after much discussion with some Norwegians, I discovered that comparatively few people read NN. So, we came up with the following fix.

• If the phone's language is Bokmål (NB-NO) take them to the NO Wikipedia.
• If the phone's language is Nynorsk (NN-NO) take them to the NN Wikipedia.

However, very few phones support NN (none have ever used QRpedia) so I'm not sure if this is the correct approach.

Others

There are lots of other languages with Wikipedia supports, but which aren't well supported on phones. Wikipedia is available in nearly 300 different languages - from Scots and Simple English to Esperanto and Latin. Although, curiously, there's no separate Wikipedia for British English - or other regional English variants, nor is there one in Klingon

The Future

So, what should QRpedia do in the future? How should it handle all the thousands of languages in conjunction with the hundreds of Wikipedia languages?

That's where you come in.

If you've got a good idea on how we handle your favourite language - drop a comment on this blog.

If you're a coder, QRpedia is open-source. Check out the code and leave a comment, or raise a bug.

We need your help to determine what we do next.

So, a few days ago, he popped down to The Children's Museum of Indianapolis to see the work the museum is doing with its Wikipedian in Residence - Lori Philips.

What else did he do while he was there? Why, scanned some QRpedia codes!

You can see all the photos of his visit on Wikimedia Commons.

My highly placed sources tell me that Mr Wales was highly impressed with the work the museum had done, and loved their innovative use of QRpedia.

If you'd like to get started with QRpedia - visit the main site to get started. It's free, fast, and fantastic!

The most common question about QRpedia is "What does it do if the article doesn't exist in my language?"

Consider the following example...

• A French user is in a German museum.
• They scan a code - which points to de.qrwp/Nelahozeves
• Unfortunately, Wikipedia doesn't have the "Nelahozeves" article in French
• What should QRpedia do?

Choices

This has been a matter for much debate in the QRpedia team. I would greatly appreciate your thoughts on the subject.

As we see it, our choices are...

1. 404 - Tell the user that the page doesn't exist.
2. Wrong Language - Show the French user the German page
3. Choice of Language - Show the French user a choice of German, English, Dutch, etc.
4. Smart Search - Search French Wikipedia for the term, display the results to the user.
5. Latin - Omnis intelligit Latinum. Si non sunt barbarus!

At the moment, we do the smart search. In this case, we point to the French Wikipedia search for Nelahozeves - which has a couple of relevant results.

This isn't ideal. Often there are no suitable results - especially as we are searching French Wikipedia for a German word or phrase.

So - over to you. What should we do? Please leave your comments and suggestions.

• QRpedia was presented by Roger at Wikimania - it seemed to go down a storm!
• The Children's Museum of Indianapolis are adding more QR codes. A lovely blog from Lori about seeing people scan the codes.
• A bug in Wikipedia has been fixed. This means that detection of mobile now happens at Wikipedia's front-end.
• The British Museum are stepping up their engagement with Wikipedia. There's an interesting quote about QRpedia from Matthew Cock

  Cock says there can be benefits in relation to the use of QR (quick response) code technology, when a QR reader on a mobile device links to a QR code on a label by an object in a gallery or museum.

  "Not only does the link take you to the Wikipedia page for that object, but it also reads which language your phone is set at and takes you to the correct language version," he explains.

  QR technology has been tested in museums, including the Derby Museum, but he says the British Museum would not pursue its use unless the code directed the user to the museum's own web pages.

A few more top-secret events in the pipeline - watch this space!

This is the video from the Derby Museum Backstage Pass where we gave the first public demonstration of QRpedia.

The Video

Video shot by Nick Moyes.

The Slides

All slides are a work in progress. This is a close approximation of what was presented on the day.

View more presentations from Terence Eden

I was invited there to talk about QRpedia and how it can improve visitor engagement in GLAM - Galleries Libraries Archives Museums.

After the introductions, I had a quick wander around the museum - where I found the first set of QRpedia codes! I was, as this video will attest, very excited!

I'll place the slides and video of the talk on the web once they're ready. In the meantime, here's what I learned from the day.

Lessons Learned

Not everything went perfectly with our first public trial - that's to be expected. So here are the top tips that we discovered from the day

Too tall

I couldn't help but laugh as a young boy struggled to scan the code right at the top of this exhibit. While watching him jump up and down was hilarious - there's a serious point about accessibility. Not everyone can reach as high as the code was placed - especially wheelchair users. It might be worth having some guidelines around code placement.

Too small

Some of the codes were printed quite small. My original intention was to have postcard sized QR codes - large enough to easily see and for several people to scan at once. What I hadn't figured on was the lack of space in the exhibits and the sheer number of codes that would be there. Here's a snap from one of the display cases, I've highlighted the codes. They were just about scanable - but I still think printing them larger will make it easier to scan. It's especially important if the code is behind glass and visitors can't get close to them.

Too Much Glare

With the light from the exhibits, the laminate on the QR codes, and the glass cases, there was a lot of glare. Luckily the QR codes were all scanable thanks to the error-correction built in. I think it's probably best not to laminate the codes - keep them as matt as possible.

Unexpected Usage

One thing I hadn't counted on was the creative uses to which the codes can be put. For example, one of the most famous paintings in Derby Museum was on loan. Rather than just put up a regular sign, they'd put up a QRpedia code so that visitors could still see the painting and read all about its history. Brilliant! Whenever you create a project, don't be surprised if people put it to a use that you had never envisioned.

Code Design

As well as the technical aspects of code design - making them large enough and with sufficient border space - there are some practical considerations. Without any human-readable identification, it's hard for museum staff to place the codes accurately! A big pile of QR codes are easy to mix up - and finding which code goes with which exhibit can be tricky.

Too wobbly

Rather embarrassingly, I tried to demonstrate how easy it is to scan in QR codes using a code printed on a free-standing poster. The breeze in the room was sufficient to cause the poster to billow back-and-forth meaning my camera couldn't focus on it. QRpedia codes must be afixed to something solid for best results.

The Language Issue

What do you do if an article isn't available in the visitor's preferred language? My initial suggestion was not met with universal acclaim.

Andy Mabbett

@pigsonthewing

#GLAMDerby "Default to Latin", quips @edent

---

Hey, it's not my fault they don't teach Latin any more! The suggestions which came out of the discussion were...

• Default to the article in the language of the museum.
• Redirect the visitor to the Mobile Wikipedia homepage in their preferred language - then they can search for something similar.
• QRpedia should look through the article and try to find something similar in the visitor's language.
• Present a screen showing all the available languages for that article. Let the user choose which language to read. This is what QRpedia does currently.

There are no easy answers - I guess we'll have to do some trials to see what people prefer. Any suggestions welcomed!

IE6

Bane of web developers everywhere. The accursed IE6 is still in use in museums who can't afford to upgrade their computers. QRpedia.org only works in modern browsers. So we're going to have to do some redesigning to make sure that everyone can use the service.

...And Finally

After an amazing day, we retired to the The Old Silk Mill pub. There, I found my new favourite stout - Spire's Sgt Pepper. Made with ground black pepper. it provided the perfect end to a perfect day. Thanks to everyone who made the day such a success - especially the gloriously bearded Andy Mabbett, the hugely entertaining Richard Mackney, Wikimedia's Andrew Turvey for his kind offer of help, the knowledgeable Hannah Fox, Nick Moyes for all his hard work during the day, the BBC's Mark Ansell for interviewing me about QRpedia, Marrianne Bamkin for pointing me back towards the station, and - of course - Roger Bamkin without whom none of this would have happened.

A special thank you to all the staff and volunteers at Derby Museum for making us feel so welcome - and for allowing us to pepper their exhibits with our QR codes.

Today Derby - tomorrow the world!

There are five key points to the success of such a scheme.

1. 100% of visitors will be scanning these codes on their mobile phones. The code must point to the mobile version of Wikipedia.
2. <100% of visitors will speak the language of the country where the museum is located (for example, 25% of visitors to the Science Museum are foreign. Pointing to just the English article is unacceptable.
3. The QR code should be simple enough to scan quickly. This mean using as short a URL as possible.
4. The URL presented must be human readable. The user must know to where they are being directed. The QR scanner may have a "history" option which will allow the user to see the codes they've scanned.
5. Statistics must be gathered showing how many people are using the QR codes to assess their effectiveness.

Mobile Redirection is Broken on Wikipedia

The current way that Wikipedia does mobile redirection is broken and should be considered harmful to users. At the moment, a mobile user-agent visiting http://en.wikipedia.org/wiki/Rossetta_Stone has to load the entire HTML document, process the complex JavaScript, and only then is redirected to the mobile version of the site.

This means that a visitor incurs a significant wait before they are directed to the correct content. If the visitor is roaming in a foreign country, they may pay significant per-MB costs for this download.

The correct way is for Wikipedia's servers to detect the mobile user-agent and 302 redirect to the mobile version.

Language Detection

I don't know how Wikipedia detects the preferred language of its visitors. Ideally, it should look at the Accept-Language Header of the phone and / or use the IP address of the device (assuming the user isn't using the museum's WiFi).

Finally, the QR/mobile version of Wikipedia should allow a user to easily change the language of the page they are viewing.

Short URL, Human Readability, and Statistics

URL shortening services often produce a jumble of letters and number which, while short, mean nothing to the human user. For example http://bit.ly/fNXn1W. In addition, use of commercial URL shortening services is problematic should the company no longer maintain its service.

Using the "normal" URL gives us this

Using the "mobile" URL gives us this

Using a QR code specifically for QR use gives us this

Finally, there could be a separate short domain for Wikipedia accessed through QR codes. For example, the currently unregistered qrpedia.org Obviously, this has branding problems - would people recognise it as being from Wikipedia?

Conclusion

Ideally, we would use a URL like http://wikipedia.org/qr/Rossetta_Stone

When the use scanned the code and requested the URL, Wikipedia would then perform two actions

1. Detect the user's preferred language.
2. 302 redirect to the mobile version.

If the language detection was incorrect, the user could simply change it.

Wikipedia admins, museum staff, and anyone else, would be able to see which exhibits had the most scans.

Thoughts?