How to translate a game into 20 languages and avoid going to hell
Greetings! I’m Vladimir Kupratsevich, head of translation projects at Alconost.
We were once approached by a client – a game developer for Android and iOS. He had a game he wanted localized into 20 languages, including Arabic, Japanese, and Korean.
Why not? we thought. We know how to localize apps and we love doing it; we have the native-speaker translators for it. But the devil was in the details.
Getting ahead of myself, let me say that we successfully performed the exorcism, and gained some useful experience in the bargain. We decided to conduct the translations on a cloud platform – and we were not mistaken.
Now let me tell you about the devils we exorcised using this solution. I suspect every developer and translator who works in IT can learn something useful from our experience.
The first devil
The game is updated regularly. On average, each day three lines need to be translated into 20 languages. This means we can’t simply deliver a turn-key translation and pack all the project documentation away into a trunk with a sigh of relief.
For the new lines may refer to subjects and concepts already contained in the game. They need to be translated consistently, so that a Scarecrow once translated as such does not appear as a Bogeyman or an Uglie-Wuglie.
Exorcising the first devil: First, a custom terminology database is created. We immediately acquaint each new translator joining the project with this glossary (if one of the “Big Twenty" transfers to another project or drops out for other reasons). Second, this database is always provided to the proofreader – the editor who reads through the translated text.
Benefit for the client: The result is that the client receives quality translations into the required languages. We avoid bothering the client by asking the same questions repeated by different associates: the translator’s access to the glossary immediately eliminates a number of questions. And quality control by the editors is the best defense against less-than-ideal interpretations, inaccuracies, and typos.
The second devil
Frequently, identical phrases are encountered in versions of the app for iOS and Android. To translate each phrase separately would be expensive and impractical. To hunt for the equivalent in the localization for the other platform would be inconvenient and time-consuming.
Exorcising the second devil: The cloud platform we employ contains a “translation memory" option. If word sequences or phrases in the project have previously been translated (even for a different OS), the system automatically supplies the existing translation. Like all phrases taken from the translation memory, the phrase is invariably checked by an editor, who notifies the translator if for some reason the auto-suggested version is not a good match.
Benefit for the client: Localization cost is ordinarily calculated by the number of symbols. Having a translation memory in the cloud system reduces the amount of text for “live" translation. This means a lower cost of localization for the client.
Even if the localization is only for one OS, a translation memory still comes in handy for auto-suggesting standard phrases. For example, if we begin the first quest with the phrase: “Great things await you, O young Padawan!", the second quest should begin the same way, so as not to confuse the user.
Translation memory not only optimizes the client’s expenses, but also helps ensure consistency of the text, to help users feel confident and well-oriented in the game.
The third devil
There are 20 translators and 20 editors working on a translation into 20 languages. That makes 40 people, not counting the manager. These talented people live in various time zones. But this factor needs to have no negative impact on the release date.
Ideally, the moment the developers enter new phrases into the text files these phrases should immediately be submitted to each of the 20 translators, so that the latter can commence localization without delay. The developers have the same expectations: ideally, the translated lines should immediately appear in the project the moment the editor has finished checking them. “Immediately" means that very instant, not some time tomorrow.
If delays occur at any of these stages – whether it be the “incoming" lines for translation or “outgoing" delivery of localized texts – we go hurtling headlong into a cauldron of boiling pitch.
Exorcising the third devil: We used the cloud platform’s file API to automate text delivery from and submission of completed texts to the client. This meant we could toss 20 task assignment emails to translators into the burn pile, and as many emails to editors. Task assignment occurs automatically: the completed translation is immediately sent for editing, and the localized texts are instantly incorporated into the project. Of course, there is the risk that a translator or editor may ignore an assigned task, but our fellows have never been caught slacking, God love ’em. In any case, the manager is always on the lookout, keeping an eye on how work progresses to make sure the translations into all the languages proceed with equal speed.
Benefit for the client: The client saves time, and the risk of a delayed release due to non-readiness of the localized versions is minimized.
The fourth devil
What’s to be done with a phrase if the context of its usage is unclear? Should a humorous phrase – a meme, for example – be translated literally, or creatively localized? The client sends three phrases in all caps, and three in all lower-case letters. Should this be preserved in the localized version?
These and 100 other questions are a living hell for the manager, with every translator sending his own list of comments. To ask the client all these questions would mean to inundate him with dozens of e-mails and bring work to a grinding halt. Yet to not ask them is out of the question.
Exorcising the fourth devil: The cloud platform we employ provides a place for discussion of each phrase. The translator and the client discuss ambiguous questions directly, avoiding the manager middleman, which helps to keep up the pace of the work. The manager is spared from drowning in a muddle of questions, mixing up questions from the Japanese translator with comments from the French localizer. What’s not to like?
Benefit for the client: The work process is completely transparent for the client. By being directly in contact with the translators regarding ambiguities, the client can effectively influence the end result and immediately select the versions he likes best. This translates to fewer editing cycles, as all corrections are made in the course of the work.
Read more about these and other devils in my presentation, “How We Made Localization Agile".
It is my hope that our experience will help you conduct localization without the headache. If you don’t have time to spend looking for translators and management, look us up. We’ve got it all.