Archive for the 'Wikipedia' Category

The Curious Problem of Belarusian and Igbo in Twitter and Bing Translation

Twitter sometimes offers machine translation for tweets that are not written in the language that I chose in my preferences. Usually I have Hebrew chosen, but for writing this post I temporarily switched to English.

Here’s an example where it works pretty well. I see a tweet written in French, and a little “Translate from French” link:

Emmanuel Macron on Twitter.png

The translation is not perfect English, but it’s good enough; I never expect machine translation to have perfect grammar, vocabulary, and word order.

Now, out of curiosity I happen to follow a lot of people and organizations who tweet in the Belarusian language. It’s the official language of the country of Belarus, and it’s very closely related to Russian and Ukrainian. All three languages have similar grammar and share a lot of basic vocabulary, and all are written in the Cyrillic alphabet. However, the actual spelling rules are very different in each of them, and they use slightly different variants of Cyrillic: only Russian uses the letter ⟨ъ⟩; only Belarusian uses ⟨ў⟩; only Ukrainian uses ⟨є⟩.

Despite this, Bing gets totally confused when it sees tweets in the Belarusian language. Here’s an example form the Euroradio account:

Еўрарадыё   euroradio    Twitter double.pngBoth tweets are written in Belarusian. Both of them have the letter ⟨ў⟩, which is used only in Belarusian, and never in Ukrainian and Russian. The letter ⟨ў⟩ is also used in Uzbek, but Uzbek never uses the letter ⟨і⟩. If a text uses both ⟨ў⟩ and ⟨і⟩, you can be certain that it’s written in Belarusian.

And yet, Twitter’s machine translation suggests to translate the top tweet from Ukrainian, and the bottom one from Russian!

An even stranger thing happens when you actually try to translate it:

Еўрарадыё   euroradio    Twitter single Russian.pngNotice two weird things here:

  1. After clicking, “Ukrainian” turned into “Russian”!
  2. Since the text is actually written in Belarusian, trying to translate it as if it was Russian is futile. The actual output is mostly a transliteration of the Belarusian text, and it’s completely useless. You can notice how the letter ⟨ў⟩ cannot be transliterated.

Something similar happens with the Igbo language, spoken by more than 20 million people in Nigeria and other places in Western Africa:

 4  Tweets with replies by Ntụ Agbasa   blossomozurumba    Twitter.png

This is written in Igbo by Blossom Ozurumba, a Nigerian Wikipedia editor, whom I have the pleasure of knowing in real life. Twitter identifies this as Vietnamese—a language of South-East Asia.

The reason for this might be that both Vietnamese and Igbo happen to be written in the Latin alphabet with addition of diacritical marks, one of the most common of which is the dot below, such as in the words ibụọla in this Igbo tweet, and the word chọn lọc in Vietnamese. However, other than this incidental and superficial similarity, the languages are completely unrelated. Identifying that a text is written in a certain language only by this feature is really not great.

If I paste the text of the tweet, “Nwoke ọma, ibụọla chi?”, into translate.bing.com, it is auto-identified as Italian, probably because it includes the word chi, and word that is written identically happens to be very common in Italian. Of course, Bing fails to translate everything else in the Tweet, but this does show a curious thing: Even though the same translation engine is used on both sites, the language of the same text is identified differently.

How could this be resolved?

Neither Belarusian nor Igbo languages are supported by Bing. If Bing is the only machine translation engine that Twitter can use, it would be better to just skip it completely and not to offer any translation, than to offer this strange and meaningless thing. Of course, Bing could start supporting Belarusian; it has a smaller online presence than Russian and Ukrainian, but their grammar is so similar, that it shouldn’t be that hard. But what to do until that happens?

In Wikipedia’s Content Translation, we don’t give exclusivity to any machine translation backend, and we provide whatever we can, legally and technically. At the moment we have Apertium, Yandex, and YouDao, in languages that support them, and we may connect to more machine translation services in the future. In theory, Twitter could do the same and use another machine translation service that does support the Belarusian language, such as Yandex, Google, or Apertium, which started supporting Belarusian recently. This may be more a matter of legal and business decisions than a matter of engineering.

Another thing for Twitter to try is to let users specify in which languages do they write. Currently, Twitter’s preferences only allow selecting one language, and that is the language in which Twitter’s own user interface will appear. It could also let the user say explicitly in which languages do they write. This would make language identification easier for machine translation engines. It would also make some business sense, because it would be useful for researchers and marketers. Of course, it must not be mandatory, because people may want to avoid providing too much identifying information.

If Twitter or Bing Translation were free software projects with a public bug tracking system, I’d post this as a bug report. Given that they aren’t, I can only hope that somebody from Twitter or Microsoft will read it and fix these issues some day. Machine translation can be useful, and in fact Bing often surprises me with the quality of its translation, but it has silly bugs, too.

Advertisements

Five More Privileges of English Speakers, part 2: Language and Software

For the previous part in the series, see Five Privileges of English Speakers, part 1.

I’m continuing the series of posts in each of which I write about five privileges that English speakers have without giving it a lot of thought. The examples I give mostly come from my experience translating software, Wikipedia articles, blog posts, and some other texts between English, Hebrew, and Russian. Hebrew and Russian are the languages I know best. If you have interesting examples from other languages, I am very interested in hearing them and writing about them.

I’m writing them mostly as they come into my mind, without a particular order, but the five items in this part of the series will focus on usage of the English language in software, and try to show that the dominance of English is not only a consequence of economics and history, but that it’s further reinforced by features of the language itself.

1. Software usually begins its life in English

English is the main language of software development worldwide.

The world’s best-known place for software development is Silicon Valley, an English-speaking place. That’s the place of Facebook, Google, Apple, Oracle and many others. California is also the home of Adobe.

There are several other hubs of software development in United States: Seattle (Microsoft, Amazon), North Carolina (Red Hat), New York (IBM, CA), Massachusets (TripAdvisor, Lotus, RSA), and more. The U.S. is also the source for much of computer science research and education, coming from Berkeley, MIT, and plenty of other schools. The U.S. is also the birthplace of the Internet, originally supported by the U.S. Department of Defense and several American universities. The world wide web, which brought the Internet to the masses, was created in Switzerland by an English speaker.

Software is developed in other countries—India, Russia, Israel, France, Germany, Estonia, and many other countries. But the dominance of the U.S. and of the English language is clear. The reason for this is not only that the U.S. is the source for much of computer technologies, but also—and probably more importantly—that the U.S. is the biggest consumer market for software. So developers in all countries tend to optimize the product for the highest-paying consumers, and these only need English.

When engineers write the user interface of their software in English, they often do not give any thought to other languages at all, or make translation possible, but complicated by English-centric assumptions about number, gender, text direction, text size, personal names, and plenty of other things, which will be explored in further points.

2. Terminology

English is also the source for much of the computer world’s terminology. Other languages have to adapt terms like smartphone, network, token, download, authentication, and thousands of others.

Some language communities work hard to translate them all meticulously into native words; Icelandic, Lithuanian, French, Chinese, and Croatian are famous examples. This is nice, but requires effort on behalf of terminology committees, who need to keep up with the fast pace of technological development, and on behalf of the software translators, who have to keep with the committees.

Some just transliterate most of them: keep the term essentially in English, but rewritten in the native alphabet. Hindi and Japanese are examples of that. This seems easy, but it is based on a problematic assumption: that the target language speakers who will use the software know at least some English! This assumption is correct for the translators, who don’t just know the English terms, but are probably also quite accustomed to it, but it’s not necessarily correct for the end users. Thus, the privilege is perpetuated.

Some languages, such as Hebrew, German, and Russian, are mid-way, with language academics and purists pulling to purer native language, engineers pulling to more English-based words, and the general public settling somewhere in between—accepting the neologisms for some terms, and going for English-based words for others.

For the non-English languages it provides fertile ground for arguments between purists and realists, in which the needs of the actual users are frequently forgotten. All the while, English speakers are not even aware of all this.

3. Easy binary logic word formation

One particular area of computer terminology is binary logic. This sounds complicated, but it’s actually simple: in electronics and software opposite notions such as true / false, success / failure, OK / Cancel, and so forth, are very common.

This translates to a great need for words that express opposites: enable / disable, do / undo, log in / log out, delete / undelete, block / unblock, select / deselect, online / offline, connect / disconnect, read / unread.

Notice something? All of the above words are formed with the same root, with the addition of a prefix (un-, dis-, de-), or with the words “on” and “off”.

A distinct, but closely related need, is words for repetition. Computers are famously good at doing things again and again, and that’s where the prefix re- is handy: reconnect, retry, redo, retransmit.

These features happen to be conveniently built into the English language. While English has extremely simple morphology for declension and conjugation (see the section “Spell-checking” in part 1 of the series), it has a slightly more complex morphology for word formation, but it’s still fairly easy.

It is also productive. That is, a software developer can create new words using it. For example, the MediaWiki software has the concept of “oversight”—hiding a problematic page in such a way that only users with a particular permission can read it. What happens if a page was hidden by mistake? Correct: “unoversight”. This word doesn’t quite exist elsewhere, but it doesn’t sound incorrect, because familiar English word formation rules were used to coin it.

As it always happens, English-speaking software engineers either don’t think about it at all, or think that other languages also have similar word formation rules. If you haven’t guessed it already, it is not true. Sime other European languages have similar constructs, but not necessarily as consistent as in English. And for Semitic languages like Hebrew it’s a disaster, because in Semitic languages prefixes are used for entirely different things, and the grammar doesn’t have constructs for repetition and negation. So when translating software user interface strings into Hebrew, we have to use different words as opposites. For example the English pair connect / disconnect is translated as lehitḥabér / lehitnaték—completely different roots, which Hebrew is just lucky to have. Another option is to use negative words like lo and bilti, or bitul, but they are often unnatural or outright wrong. Having to deal with something like “Mark as unread” is every Hebrew software translator’s nightmare, even though it sounds pretty straightforward in English.

English itself also has pairs of negative words that are not formed using the above prefixes, for example next / previous and open / close, but in many other languages they are much more common.

4. Verbing

“Verbing weirds language”, as one of the famous Calvin and Hobbes panels says.

Despite being a funny joke in the comic, it’s a real feature of the English language: because of how English morphology and syntax work, nouns can easily jump into the roles of adjectives and verbs without changing the way they are written.

For English, this is a useful simplification, and it works in labeling, as well as in advertising. “Enjoy Coca-Cola” is something more than an imperative. The fact that it’s a short single word and that it’s the same in all genders and numbers, makes it more usable as a call to action than it would be in other languages. And, other than advertising, where are calls to action very common? Software, of course. When you’re trying to tell a user to do something, a word that happens to be both the abstract concept and the imperative is quite useful.

Perhaps the most famous example of this these days is Facebook’s “Like”. Grammatically, what is it in English? Imperative? A noun describing an abstract action? Maybe a plain old noun, as in “chasing likes” (this is a plural noun—English verb don’t have a plural form!)? Answer: it’s all of them and more.

When translated to Hebrew in Facebook’s interface, it’s Ahávti, which literally means “I loved it”. Actually, this translation is mostly good, because it’s understandable, idiomatic, and colloquial enough without compromising correctness. Still, it’s a verb, which is not imperative, and it’s definitely not a noun, so you cannot use it in a sentence as if it was a noun. Indeed, Hebrew speakers are comfortable using this button, but when they speak and write about this feature, they just use its English name: “like” (in plural láykim). It even became a slightly awkward, but commonly used verb: lelaykék. Something similar happens in Russian.

It would be impossible in Hebrew and Russian to use the exact same word for the noun and the verb, especially in different persons and genders. Sometimes the languages are lucky enough to be able to adapt an English verb in a way that is more or less natural, but sometimes it’s weird, and hurts the user experience.

5. Word length

This one is relatively simple and not unique to English, but should be mentioned anyway: English words are neither very long, nor very short. Examples of languages where words are, on average, longer than in English, are Finnish, Tamil, German, and occasionally Russian. Hebrew tends to be shorter, although sometimes a single English word has to be translated with several Hebrew words, so it can get also get longer. This is true for a pretty much any language, really.

In designing interfaces, especially for smaller screens, the length of the text is often important. If a button label is too long, it may overflow from the button, or be truncated, making the display ugly, or unusable, or both.

If you’re an English speaker, it probably won’t happen with you, because almost all software is usually designed with the word length of your language in mind. Other languages are almost always an afterthought.

The good practice for software engineers and designers is to make sure that translated strings can be longer. Their being shorter is rarely a problem, although sometimes a string is so short that the button may become to small to click or tap conveniently.


Generally, what can you do about these privileges?

Whoever you are, remember it. If you know English, you are privileged: Software is designed more for you than for people who speak other languages.

If you are a software engineer or a designer, at the very least, make your software translatable. Try to stick to good internationalization practices and to standards like Unicode and CLDR. Write explanations for every translatable string in as much detail as possible. Listen to users’ and translators’ complaints patiently—they are not whining, they are trying to improve your software! The more internationalizable it is, the more robust it is for you as a developer, and for your English-speaking users, too, because better design thinking will be going into each of its components, and less problematic assumptions will be made.

Five Privileges of English Speakers, part 1

It’s very common today on progressive blogs to urge people to check their privilege.

Being an English speaker, native or non-native, is a privilege.

It’s not as often as discussed as other forms of privilege, such as white, male, cis, hetero, or rich privilege. The reason for this is simple: The world’s media is dominated by the English language. English-language movies are more popular in many countries than movies in these countries’ own languages, English-language news networks are quoted by the rest of the world, the world’s most popular social networks are based in the U.S. and are optimized for U.S. audiences, etc.

So, when English speakers discuss privilege among each other, English is not much of an issue, and they dedicate more time to race, gender, wealth, religion, and other factors that differentiate between people in English-speaking countries.

Despite this, I am not the first one to describe English as a privilege. A simple Google search for english language privilege will yield many interesting results.

What I do want to try to do in this series of posts is to list the particular nuances that make English such a privilege in as much detail as possible. I wanted to write this for a long time, but there are many such nuances, so I’ll just do it in batches of five, in no particular order:

1. Keyboard

If you speak English, congratulations: A keyboard on which your language can be written is available on all electronic devices.

All of them.

All desktops, laptops, phones, tablets, watches. The only notable exception I can think of is typewriters, which only makes the point more tragic: technology moved forward and made writing easier in English, but harder in many other languages, where local-language typewriters were replaced with computers with English-only keyboard.

At the very worst case, writing English on a computer will be slightly inconvenient in countries like Germany, France, or Turkey, where the placement of the Latin letters on the keys is slightly different from the U.S. and U.K. QWERTY standard. Oh, poor American tourists.

On a more serious note, though, even though a lot of languages use the Latin alphabet, a lot of them also use a lot of extra diacritics and special characters, and English is one of the very few that doesn’t. Of the top 100 world’s languages by native speakers, only Malay, Kinyarwanda, Somali, and Uzbek have standardized orthographies that can be written in the basic 26-letter Latin alphabet without any extra characters. We can also add Swahili, which has a large number of non-native speakers, but that’s it. With other languages you can get stuck and not be able to write your language at all (Hindi, Chinese, Russian, etc.), or you may have to write in a substandard orthography because you can’t type letters like é or ł (French, Vietnamese, Polish, etc.).

The above is just the teeny-tiny tip of the iceberg; the keyboard problem will be explored in more points later.

2. Spell-checking

English word morphology is laughably simple.

There’s -s for plurals and for third person present tense verbs, there’s -‘s for possession, and there are -ed and -ing verb forms. There are also some contractions (‘d, ‘s, ‘ll, ‘ve), and a long, but finite list of irregular verb forms, and an even shorter list of irregular plural noun forms. And that’s it.

Most languages aren’t like that. In most languages words change with prefixes, suffixes, infixes, clitics, and so on, according to their role in the sentence.

Beyond the fact that English writing is (arguably) easier for children and foreigners to learn, this means that software tools for processing a language are easy to develop for English and hard to develop for other languages.

The first simple example is spell-checking.

English has had not just spelling, but also grammar and style checkers built into common word processors for decades, and many languages of today don’t even have spelling checkers, not to mention grammar, or style, or convenient searching. (See below.)

So in English, when you type “kinh”, most word processors will suggest correcting it to “king”, but then, some of them may also suggest replacing this word with “monarch” to be more inclusive for women, and this is just one of the hundreds of style improvement suggestions that these tools can make. For a lot of other languages, even simple spell-checking of single words hasn’t been developed yet, and grammar checking is a barely-imaginable dream.

3. Autocompletion

Simpler morphology has many other effects.

Even though Russian is my first native language and I speak it more fluently than I speak English, I am much slower when I’m typing in Russian on my phone. In English, the autocompleting keyboard makes it possible to write just two or three letters of a word and let the software complete the rest. In Russian, the ending of the word must be typed, and autocompletion rarely guesses it correctly. Typing an incorrect ending will make a sentence convey incorrect information, or just make it completely ungrammatical.

4. Searching

A yet-another issue of the previous point, English’s very simple morphology makes searching easier.

For example word processors have a search and replace function. For English, it will likely find all forms of the word, because there are so few of them anyway. But in Hebrew and Arabic, letters are often inserted or changed in the middle of the word according to its grammatical state, and you need to search for each form, which is quite agonizing. It’s comparable to “man” vs. “men” in English, except that in English such changes are very rare, while in many other languages it happens in almost every word.

With search engines that must find words across thousands of documents it gets even harder. Google can easily figure out that if you’re searching for “drive”, you may also be interested in “driving”, “drove”, and “driven”, but Russian has dozens of other forms for this word. A few languages are lucky: special support was developed for them in search engines, and tasks of this kind are automated, but most languages our just out in the cold. But English barely needs extra support like this in the first place.

5. Very little gender

A lot can be said about gendered language, but as far as basic grammar goes, English has very little in the area of gender. “He” and “She”, and that’s about it. There are also man/woman, actor/actress, boy/girl, etc., but these distinctions are rarely relevant in technology.

In many other languages gender is far more pervasive. In Semitic and Slavic languages, a lot of verb forms have gender. In English, the verb “retweeted” is the same in “Helen retweeted you” and “Michael retweeted you”, but in Hebrew the verb is different. Because Twitter doesn’t know that Helen needs a different verb, it uses the masculine verb there, which sounds silly to Hebrew speakers.

I asked Twitter developers about this many times, and they always replied that there’s no field for gender in the user profile. It becomes more and more amusing lately, now that it has become so common —and for good reasons!— to mention what one’s preferred pronouns are in the Twitter profile bio. So people see it, but computers don’t.

On a more practical note, in the relatively rare cases when third person pronouns must be used in software strings, English will often use the singular “they” instead of “he” or “she”. So English-speaking developers do notice it, but not as often as they should, and when they do, they just use the lazy singular-they solution, which is socially acceptable and doesn’t require any extra coding. If only they’d notice it more often, using their software in other languages would be much more convenient for people of all genders.

The only software packages that I know that have reasonably good support for grammatical gender are MediaWiki and Facebook’s software. I once read that Diaspora had a very progressive solution for that, but I don’t know anybody who actually uses it. There may be other software packages that do, but probably very few.


These are just the first five examples of English-language privilege I can think of. There will be many, many more. Stay tuned, and send me your ideas!

Wikipedia, a Jamaican Jew, and Yak Shaving

For me, writing in Wikipedia is very often a story, within a story, within a story.

I am a member of the Language committee, which examines and approves the creation of editions of Wikipedia in new languages.

Recently we approved the new edition in the Jamaican language—an English-based creole commonly heard in reggae, in which books were published, and into which “the usual suspects” were translated: The New Testament, Alice’s Adventures in Wonderland, The Little Prince—and now, Wikipedia.

Since the draft “incubator” Wikipedia in this language conformed to the requirements for creating a full-fledged new domain, I supported the domain’s creation. My work as a language committee member could end here—and I’m a volunteer there to begin with—but I nonetheless decided to shave a yak.

bos_grunniens_at_letdar_on_annapurna_circuit

Normal people, when they need a sweater, buy one in a store. I consider shaving a yak.

Some time after a Wikipedia in a new language is created, all the draft articles from the incubator are imported. When that is completed, I go over the list of imported articles and try to see whether there are any that aren’t linked to their counterparts in other languages. With some topics it’s easy by guessing the name of the topic or by looking at the images, and with some others it’s hard. With an English-based creole it’s of course very easy.

And that’s how the Jamaican Wikipedia ended up with only one article that doesn’t have a version in any other language: Aizak Mendiz Belisario.

It was easy enough to understand that this was a Jewish artist who lived in Jamaica in the 19th century. He was already mentioned a couple of times in the English Wikipedia, but there was no whole article about him. So I thought: Jamaican is similar enough to English and I can understand what most of the article is about, and the artist seems notable enough for an encyclopedia, because he was one of the pioneers of art in Jamaica, and because an anthology about him was published recently. And, of course, I am in a team that develops Content Translation—a translation tool for Wikipedia articles. So I decided to translate it to English.

As soon as I started the translation process, I noticed a bug. So I filed it, and because it was so easy to fix, I just fixed it.

Then I started actually translating the article. On the way I learned about the John Canoe festival, and added another spelling variant to the article about it in English; I verified that the book about the artist was actually published (you know, hoaxes happen), and googled for some more information about the artist with the hope of improving the English article further.

belisario3

Normal people could just say “Fine, that language looks legit, let’s start a Wikipedia in it”. But I actually had to read all the articles in it, and then write a new one, improve another one, fix a bug, and write a blog post about all of it.

So here you go: Isaac Mendes Belisario, in English.

There is a story like this one behind every one of the millions and millions of articles in Wikipedia in all of its languages.

Amir Aharoni’s Quasi-Pro Tips for Translating the Software That Powers Wikipedia

As you probably already know, Wikipedia is a website. A website has content—the articles; and it has user interface—the menus around the articles and the various screens that let editors edit the articles and communicate to each other.

Another thing that you probably already know is that Wikipedia is massively multilingual, so both the content and the user interface must be translated.

Translation of articles is a topic for another post. This post is about getting all of the user interface translated to your language, as quickly and efficiently as possible.

The most important piece of software that powers Wikipedia and its sister projects is called MediaWiki. As of today, there are 3,335 messages to translate in MediaWiki, and the number grows frequently. “Messages” in the MediaWiki jargon are strings that are shown in the user interface, and that can be translated. In addition to core MediaWiki, Wikipedia also has dozens of MediaWiki extensions installed, some of them very important—extensions for displaying citations and mathematical formulas, uploading files, receiving notifications, mobile browsing, different editing environments, etc. There are around 3,500 messages to translate in the main extensions, and over 10,000 messages to translate if you want to have all the extensions translated. There are also the Wikipedia mobile apps and additional tools for making automated edits (bots) and monitoring vandalism, with several hundreds of messages each.

Translating all of it probably sounds like an enormous job, and yes, it takes time, but it’s doable.

In February 2011 or so—sorry, I don’t remember the exact date—I completed the translation into Hebrew of all of the messages that are needed for Wikipedia and projects related to it. All. The total, complete, no-excuses, premium Wikipedia experience, in Hebrew. Every single part of the MediaWiki software, extensions and additional tools was translated to Hebrew, and if you were a Hebrew speaker, you didn’t need to know a single English word to use it.

I wasn’t the only one who did this of course. There were plenty of other people who did this before I joined the effort, and plenty of others who helped along the way: Rotem Dan, Ofra Hod, Yaron Shahrabani, Rotem Liss, Or Shapiro, Shani Evenshtein, Inkbug (whose real name I don’t know), and many others. But back then in 2011 it was I who made a conscious effort to get to 100%. It took me quite a few weeks, but I made it.

Of course, the software that powers Wikipedia changes every single day. So the day after the translations statistics got to 100%, they went down to 99%, because new messages to translate were added. But there were just a few of them, and it took me a few minutes to translate them and get back to 100%.

I’ve been doing this almost every day since then, keeping Hebrew at 100%. Sometimes it slips because I am traveling or I am ill. It slipped for quite a few months because in late 2014 I became a father, and a lot of new messages happened to be added at the same time, but Hebrew is back at 100% now. And I keep doing this.

With the sincere hope that this will be useful for translating the software behind Wikipedia to your language, let me tell you how.

Preparation

First, let’s do some work to set you up.

  • Get a translatewiki.net account if you haven’t already.
  • Make sure you know your language code.
  • Go to your preferences, to the Editing tab, and add languages that you know to Assistant languages. For example, if you speak one of the native languages of South America like Aymara (ay) or Quechua (qu), then you probably also know Spanish (es) or Portuguese (pt), and if you speak one of the languages of the former Soviet Union like Tatar (tt) or Azerbaijani (az), then you probably also know Russian (ru). When available, translations to these languages will be shown in addition to English.
  • Familiarize yourself with the Support page and with the general localization guidelines for MediaWiki.
  • Add yourself to the portal for your language. The page name is Portal:Xyz, where Xyz is your language code.

Priorities, part 1

The translatewiki.net website hosts many projects to translate beyond stuff related to Wikipedia. It hosts such respectable Free Software projects as OpenStreetMap, Etherpad, MathJax, Blockly, and others. Also, not all the MediaWiki extensions are used on Wikimedia projects; there are plenty of extensions, with thousands of translatable messages, that are not used by Wikimedia, but only on other sites, but they use translatewiki.net as the platform for translation of their user interface.

It would be nice to translate all of it, but because I don’t have time for that, I have to prioritize.

On my translatewiki.net user page I have a list of direct links to the translation interface of the projects that are the most important:

  • Core MediaWiki: the heart of it all
  • Extensions used by Wikimedia: the extensions on Wikipedia and related sites
  • MediaWiki Action Api: the documentation of the API functions, mostly interesting to developers who build tools around Wikimedia projects
  • Wikipedia Android app
  • Wikipedia iOS app
  • Installer: MediaWiki’s installer, not used in Wikipedia because MediaWiki is already installed there, but useful for people who install their own instances of MediaWiki, in particular new developers
  • Intuition: a set of different tools, like edit counters, statistics collectors, etc.
  • Pywikibot: a library for writing bots—scripts that make useful automatic edits to MediaWiki sites.

I usually don’t work on translating other projects unless all of the above projects are 100% translated to Hebrew. I occasionally make an exception for OpenStreetMap or Etherpad, but only if there’s little to translate there and the untranslated MediaWiki-related projects are not very important.

Priorities, part 2

So how can you know what is important among more than 15,000 messages from the Wikimedia universe?

Start from MediaWiki most important messages. If your language is not at 100% in this list, it absolutely must be. This list is automatically created periodically by counting which 600 or so messages are actually shown most frequently to Wikipedia users. This list includes messages from MediaWiki core and a bunch of extensions, so when you’re done with it, you’ll see that the statistics for several groups improved by themselves.

Now, if the translation of MediaWiki core to your language is not yet at 18%, get it there. Why 18%? Because that’s the threshold for exporting your language to the source code. This is essential for making it possible to use your language in your Wikipedia (or Incubator). It will be quite easy to find short and simple messages to translate (of course, you still have to do it carefully and correctly).

Getting Things Done, One by One

Once you have the most important MediaWiki messages 100% and at least 18% of MediaWiki core is translated to your language, where do you go next?

I have surprising advice.

You need to get everything to 100% eventually. There are several ways to get there. Your mileage may vary, but I’m going to suggest the way that worked for me: Complete the easiest piece that will get your language closer to 100%! For me this is an easy way to strike an item off my list and feel that I accomplished something.

But still, there are so many items at which you could start looking! So here’s my selection of components that are more user-visible and less technical, sorted not by importance, but by the number of messages to translate:

  • Cite: the extension that displays footnotes on Wikipedia
  • Babel: the extension that displays boxes on userpages with information about the languages that the user knows
  • Math: the extension that displays math formulas in articles
  • Thanks: the extension for sending “thank you” messages to other editors
  • Universal Language Selector: the extension that lets people select the language they need from a long list of languages (disclaimer: I am one of its developers)
    • jquery.uls: an internal component of Universal Language Selector that has to be translated separately for technical reasons
  • Wikibase Client: the part of Wikidata that appears on Wikipedia, mostly for handling interlanguage links
  • VisualEditor: the extension that allows Wikipedia articles to be edited in a WYSIWYG style
  • ProofreadPage: the extension that makes it easy to digitize PDF and DjVu files on Wikisource
  • Wikibase Lib: additional messages for Wikidata
  • Echo: the extension that shows notifications about messages and events (the red numbers at the top of Wikipedia)
  • MobileFrontend: the extension that adapts MediaWiki to mobile phones
  • WikiEditor: the toolbar for the classic wiki syntax editor
  • ContentTranslation extension that helps translate articles between languages (disclaimer: I am one of its developers)
  • Wikipedia Android mobile app
  • Wikipedia iOS mobile app
  • UploadWizard: the extension that helps people upload files to Wikimedia Commons comfortably
  • Flow: the extension that is starting to make talk pages more comfortable to use
  • Wikibase Repo: the extension that powers the Wikidata website
  • Translate: the extension that powers translatewiki.net itself (disclaimer: I am one of its developers)
  • MediaWiki core: the base MediaWiki software itself!

I put MediaWiki core last intentionally. It’s a very large message group, with over 3000 messages. It’s hard to get it completed quickly, and to be honest, some of its features are not seen very frequently by users who aren’t site administrators or very advanced editors. By all means, do complete it, try to do it as early as possible, and get your friends to help you, but it’s also OK if it takes some time.

Getting All Things Done

OK, so if you translate all the items above, you’ll make Wikipedia in your language mostly usable for most readers and editors.

But let’s go further.

Let’s go further not just for the sake of seeing pure 100% in the statistics everywhere. There’s more.

As I wrote above, the software changes every single day. So do the translatable messages. You need to get your language to 100% not just once; you need to keep doing it continuously.

Once you make the effort of getting to 100%, it will be much easier to keep it there. This means translating some things that are used rarely (but used nevertheless; otherwise they’d be removed). This means investing a few more days or weeks into translating-translating-translating.

You’ll be able to congratulate yourself not only upon the big accomplishment of getting everything to 100%, but also upon the accomplishments along the way.

One strategy to accomplish this is translating extension by extension. This means, going to your translatewiki.net language statistics: here’s an example with Albanian, but choose your own language. Click “expand” on MediaWiki, then again “expand” on “MediaWiki Extensions”, then on “Extensions used by Wikimedia” and finally, on “Extensions used by Wikimedia – Main”. Similarly to what I described above, find the smaller extensions first and translate them. Once you’re done with all the Main extensions, do all the extensions used by Wikimedia. (Going to all extensions, beyond Extensions used by Wikimedia, helps users of these extensions, but doesn’t help Wikipedia very much.) This strategy can work well if you have several people translating to your language, because it’s easy to divide work by topic.

Another strategy is quiet and friendly competition with other languages. Open the statistics for Extensions Used by Wikimedia – Main and sort the table by the “Completion” column. Find your language. Now translate as many messages as needed to pass the language above you in the list. Then translate as many messages as needed to pass the next language above you in the list. Repeat until you get to 100%.

For example, here’s an excerpt from the statistics for today:

MediaWiki translation stats exampleLet’s say that you are translating to Malay. You only need to translate eight messages to go up a notch (901 – 894 + 1). Then six messages more to go up another notch (894 – 888). And so on.

Once you’re done, you will have translated over 3,400 messages, but it’s much easier to do it in small steps.

Once you get to 100% in the main extensions, do the same with all the Extensions Used by Wikimedia. It’s over 10,000 messages, but the same strategies work.

Good Stuff to Do Along the Way

Never assume that the English message is perfect. Never. Do what you can to improve the English messages.

Developers are people just like you are. They may know their code very well, but they may not be the most brilliant writers. And though some messages are written by professional user experience designers, many are written by the developers themselves. Developers are developers; they are not necessarily very good writers or designers, and the messages that they write in English may not be perfect. Keep in mind that many, many MediaWiki developers are not native English speakers; a lot of them are from Russia, Netherlands, India, Spain, Germany, Norway, China, France and many other countries, and English is foreign to them, and they may make mistakes.

So report problems with the English messages to the translatewiki Support page. (Use the opportunity to help other translators who are asking questions there, if you can.)

Another good thing is to do your best to try running the software that you are translating. If there are thousands of messages that are not translated to your language, then chances are that it’s already deployed in Wikipedia and you can try it. Actually trying to use it will help you translate it better.

Whenever relevant, fix the documentation displayed near the translation area. Strange as it may sound, it is possible that you understand the message better than the developer who wrote it!

Before translating a component, review the messages that were already translated. To do this, click the “All” tab at the top of the translation area. It’s useful for learning the current terminology, and you can also improve them and make them more consistent.

After you gain some experience, create a localization guide in your language. There are very few of them at the moment, and there should be more. Here’s the localization guide for French, for example. Create your own with the title “Localisation guidelines/xyz” where “xyz” is your language code.

As in Wikipedia, Be Bold.

OK, So I Got to 100%, What Now?

Well done and congratulations.

Now check the statistics for your language every day. I can’t emphasize how important it is to do this every day.

The way I do this is having a list of links on my translatewiki.net user page. I click them every day, and if there’s anything new to translate, I immediately translate it. Usually there is just a small number of new messages to translate; I didn’t measure precisely, but usually it’s less than 20. Quite often you won’t have to translate from scratch, but to update the translation of a message that changed in English, which is usually even faster.

But what if you suddenly see 200 new messages to translate? It happens occasionally. Maybe several times a year, when a major new feature is added or an existing feature is changed.

Basically, handle it the same way you got to 100% before: step by step, part by part, day by day, week by week, notch by notch, and get back to 100%.

But you can also try to anticipate it. Follow the discussions about new features, check out new extensions that appear before they are added to the Extensions Used by Wikimedia group, consider translating them when you have a few spare minutes. At the worst case, they will never be used by Wikimedia, but they may be used by somebody else who speaks your language, and your translations will definitely feed the translation memory database that helps you and other people translate more efficiently and easily.

Consider also translating other useful projects: OpenStreetMap, Etherpad, Blockly, Encyclopedia of Life, etc. Up to you. The same techniques apply everywhere.

What Do I Get for Doing All This Work?

The knowledge that thanks to you people who read in your language can use Wikipedia without having to learn English. Awesome, isn’t it? Some people call it “Good karma”.

Oh, and enormous experience with software localization, which is a rather useful job skill these days.

Is There Any Other Way in Which I Can Help?

Yes!

If you find this post useful, please translate it to other languages and publish it in your blog. No copyright restrictions, public domain (but it would be nice if you credit me and send me a link to your translation). Make any adaptations you need for your language. It took me years of experience to learn all of this, and it took me about four hours to write it. Translating it will take you much less than four hours, and it will help people be more efficient translators.

Continuous Translation and Rewarding Volunteers

In November I gave a talk about how we do localization in Wikimedia at a localization meetup in Tel-Aviv, kindly organized by Eyal Mrejen from Wix.

I presented translatewiki.net and UniversalLanguageSelector. I quickly and quite casually said that when you submit a translation at translatewiki, the translation will be deployed to the live Wikipedia sites in your language within a day or two, after one of translatewiki.net staff members will synchronize the translations database with the MediaWiki source code repository and a scheduled job will copy the new translation to the live site.

Yesterday I attended another of those localization meetups, in which Wix developers themselves presented what they call “Continuous Translation”, similarly to “Continuous Integration“, a popular software deployment methodology. Without going into deep details, “Continuous Translation” as described by Wix is pretty much the same thing as what we have been doing in the Wikimedia world: Translators’ work is separated from coding; all languages are stored in the same way; the translations are validated, merged and deployed as quickly and as automatically as possible. That’s how we’ve been doing it since 2009 or so, without bothering to give this methodology a name.

So in my talk I mentioned it quickly and casually, and the Wix developers did most of their talk about it.

I guess that Wix are doing it because it’s good for their business. Wikimedia is also doing it because it’s good for our business, although our business is not about money, but about making end users and volunteer translators happy. Wikimedia’s main goal is to make useful knowledge accessible to all of humanity, and knowledge is more accessible if our website’s user interface is fully translated; and since we have to rely on volunteers for translation, we have to make them happy by making their work as comfortable and rewarding as possible. Quick deployments is one of those things that provide this rewarding feeling.

Another presentation in yesterday’s meetup was by Orit Yehezkel, who showed how localization is done in Waze, a popular traffic-aware GPS navigator app. It is a commercial product that relies on advertisement for revenue, but for the actual functionality of mapping, reporting traffic and localization, it relies on a loyal community of volunteers. One thing that I especially loved in this presentation is Orit’s explanation of why it is better to get the translations from the volunteer community rather than from a commercial translation service: “Our users understand our product better than anybody else”.

I’ve been always saying the same thing about Wikimedia: Wikimedia projects editors are better than anybody else in understanding the internal lingo, the functionality, the processes and hence – the context of all the details of the interface and the right way to translate them.

Link Wikipedia Articles in Different Languages

OK THIS IS AWESOME, and “awesome” is not a word that I use lightly.

As a gift for the second birthday of the Wikidata project, nice people at Google created a tool that helps people link articles in different languages that are not linked yet. They prepared a list with thousands of pairs of articles in different languages that are supposed to be about the same subject according to their automatic guesswork. The tool only shows such articles, and a human editor must check whether they actually match, and if they do—make the linking automatically.

There were thirty six such articles for the Hebrew–English pair. About four of them were unrelated, and I fixed the linking between the rest of them. Some of them required manual intervention, because there were interfering links to unrelated subjects. For some simple cases it took me just a few seconds, and for a few complicated ones—a few minutes.

I also tried doing the same for Russian–English, but there are over a thousand article pairs there, so I only did a few. I also did a few for Catalan and Greek, and I finished all ten pairs for Bengali, even though I don’t actually know Greek or Bengali. I just used a bit of healthy intuition and Google Translate, and I’m pretty sure that I did it well.

You can help!

Here are my suggested instructions for doing this.

Preparation:

  1. Log in to mediawiki.org. This account is used also for the tool.
  2. Now go to the tool’s site. Click Login, and allow the tool to use your mediawiki.org account.
  3. Go to settings, and choose your pair of languages.
  4. Go to “Check by list” and you’ll see a list of article pairs. If there are no suggested article pairs for the language pair you selected, go back to number 3 choose some other languages. As I wrote above, from my experience, you don’t need to know a language thoroughly to perform this useful work ;)

Now click a link to a pair of articles that looks reasonable. Articles in both languages will open side by side.

  1. If the articles are definitely not about the exact same subject, click “No” in the list and find another pair.
  2. If the articles are about the same subject and one of them doesn’t have any interlanguage links, click “Add links” in the interlanguage area. In the box that will open, write the language name of the other language in the first field and the title of the article in the other field, and then click the “Link with page” button. A list of articles in other languages will be shown. If it looks reasonable, click “Confirm”, and then “Close dialog and reload page”. That’s it, the pages are linked! Click “Yes” in the list in the linking tool and proceed to another article pair.
  3. If the articles are about the same subject, but both of them appear to have links to other language, it’s possible that explicit interlanguage links are written in the source code of the articles. To resolve this, do the following:
    1. Open both articles for editing in source mode.
    2. Scroll all the way down and find whether they have explicit interlanguage links.
    3. If these are correct links to articles about the same subjects in other languages, go to those articles, and link them using Wikidata. Note that it often happens in such cases that these are links to redirects, so the actual current title may be different.
    4. If these are links to articles about other subjects, even if they are related, remove those links. For example, if the article in Bengali is about an island, and the article in Dutch is about a city on that island, remove the link – these subject are distinct enough. Ditto if the article in English is about an American human rights organization and the article in French is about a French human rights organization.
    5. If you were able to remove all the explicit links from the source, go back to point 2 above and link the articles using Wikidata.
    6. If it’s too complicated to remove these links for any reason, feel free to go to another article, but it would be nice to leave a note about this on the articles’ talk pages so that other editors would clean this up some time.

That’s it. It may get a tad complicated for some cases, but if you ask me, it’s a lot of fun.


Archives