Personally more important, this was my first trip to China in 21 years. My first localization agency — indeed, the first business I ever started — was a high-tech Chinese->English house in Taiwan in the 1980′s. I spent an extra ten days in the Shanghai/Suzhou area re-acquiring my Chinese (details here). I also had a chance to try out technical communications from behind China’s Great Firewall (see here, recommended for anyone new traveling to China).

The participants at the LISA conference included some of the largest translation houses in the world, as well as small specialty houses working only in Chinese and more than a few major clients (partial list is here).

The setting (a five star hotel in Suzhou, one hour from Shanghai) was excellent. The three of us presenting for video games were welcomed with open arms, as there is much interest in our specialty from those outside our business. I would strongly encourage more outreach by video game localizers to organizations like LISA, with their broad background in different businesses and translations tools. They have much to offer.

My presentation (available below in Chinese and English) is a general backgrounder in video game business and history, along with the current status of the world video game industry. At the end, I piece China into the puzzle.

Conclusion? There will be less translation into Chinese (due to market particulars, including legal barriers and rampant piracy), and more work for Chinese to English and other languages (given the high-quality of locally produced MMO, Free-to-Play titles — an area where Chinese developers have a 5 year lead on the West).

Victor Alonso Lion from Pink Noise, who gave a simply excellent Spanish Language presentation at GDC last year, gave an equally excellent presentation on the differences in handling video game localizations to other kinds of localization, aimed at professionals.

The differences? Size — the latest batch of MMO’s are running over a million words per title. Malleability — no other localization branch has to swap genders, objects and actions on-the-fly like interactive entertainment. Cultural issues — different countries have strong laws or customs about what can be said or shown in entertainment.

And the list goes on …

His presentation drew the second-highest ratings at the LISA conference amongst the workgroup sessions.

LISA ASIA Forum – Localizing Interactive Media & Game

View more presentations from Victor Alonso Lion.

Rolf Klischewski not only gave a presentation, No Man’s Tool, detailing the inadequacies of the current generation of translation tools for video game work; he also ran a day-long workshop on game localization issues for professionals with little or no experience in the field.

Rolf’s presentation is available here.

Like most of the Americans in the audience, I did not know anything about Buzz. Buzz is a highly successful series of quiz games developed by Relentless Software for the PlayStation line of consoles and handhelds. The wild MC and writing immediately remind me of the “You Don’t Know Jack” series from Berkeley Systems, which I fondly remember from ten years ago. Only Buzz is much bigger, wilder and funnier.

If you haven’t played Buzz, at least look at look at this video to get a taste of it.

The localization issue here is immense. Some of the questions work well between countries (Elvis, DeLorean cars and science fiction movies translate well). However, some topics do not (Golf is popular almost wholly in English speaking countries, for instance). The problem here becomes one of not translation, but transcreation. Substantial parts of the game must created anew for the target countries.

How big a job is this? For some game types, transcreation might not be so difficult. For something as far ranging and difficult to quantify as a quiz show, however, the difficulty is extremely high. Fifteen years ago, when I first entered the game industry as a writer I was a member of the Hollywood Writers’ Guild (WGA). Of their members, the very highest paid writers in all of Hollywood were the question writers for Jeopardy, a quiz show. Jeopardy is a popular quiz program in America which has struggled in every attempt to take it outside of the US. It currently only shows in the US, Canada, Sweden, Slovakia and Russia.

Excellent and informed writing, in each language, is essential. Without it, the whole project flounders.

Target languages for Buzz are French, Italian, German, Spanish, Portuguese, Dutch, Danish, Finnish, Norwegian, Swedish, Polish and Russian. In each language, there is a team of ten including four dedicated writers — people to create original text for their countries.

Questions are divided into those suitable for an international audience and those that are not. Those that aren’t are further divided into those that require additional assets and those that do not. Everything is handled in a single CMS: Scribe.

The results are fabulous. A well reviewed, well liked game that appears to travel over borders seamlessly. Millions upon million sold. 16 total titles, numerous awards. This is a case study of a great success, in a field where failure is far more typical.

Screenshots from the presentation at GDC are below. Please click on them to see a full-sized, readable version of each.

(A Japanese version of the presentation is available here.)

Now, let’s switch gears. Imagine you have an MMO on your project list. These are big. The smallest one that Apogee has bid on was just over 500,000 words. And they have to be geared for on-the-fly changes with player characters, NPC’s, weapons/items, locations, quests … Dang, that’s a big job.

Now let’s up the ante. Let’s make it not one MMO, but SEVEN. Welcome to Sony Online Entertainment. Scroll down and take a good look at the first slide of David Kim’s presentation. Seven MMO titles (plus unannounced projects), the main SOE user website, two trading card games, ~2.0 million words every year of English just in updates … Everquest 2 alone boasts over 9 million words in its current iteration. This is a massive undertaking.

David Kim, left, of Sony Online Entertainment takes questions

David, SOE’s Localization Producer, gave the presentation at GDC in San Francisco. A few days later I had the opportunity to visit their offices in San Diego. There I met David, Raymond Nguyen (Operations Manager) and the head of International Operations, Bob McEntee. Any mistakes in this summary are mine, not theirs.

Given the crushing weight of these wordcounts, the first impression visiting their offices is, “Where is everybody?” I’ve worked for operations that handle less than 10% of SOE’s annual wordcount with much larger localization staffs. The two dozen souls at SOE take up only a part of one floor. They’re even more impressive when you realize that they also handle the actual operation of SOE’s international servers, a responsibility not normally burdened on localization teams.

Brutal, complete organization is the answer. Ad hoc translation techniques and workflows would only cause delay and inevitable quality problems, overloading the small staff.

-> Data is standardized for both content and context.
-> Translation memory at SOE holds more than 60 million words, separated by title.
-> A translation engine is completely integrated into the game.
-> English strings are assigned persistent and unique ID’s.
-> Changes to English text are automatically tracked in the system.
-> A feedback loop is installed, allowing the team to improve with time.
-> The team at SOE holds a patent in on-the-fly translation methods.

Some 40% of all translation can be handled internally using the tools developed here, improving quality and saving Sony hundreds of thousands of dollars each year in external translation costs.

One downfall with having so large a workload is the language basket and geographic availability for their titles remains small. SOE titles are translated into French, German and Japanese with only a few new expansion languages. Taking older titles into a new language (see reference above for Everquest 2′s nine million words) is a massive undertaking.

As the videogame industry evolves, publishers will begin to look more and more like the streamlined operation at SOE. Localization tools will be integrated at the very beginning of game development, automated tool sets will allow easy import and export of text and other assets, English (or other first language) changes will automatically be flagged. The penalty for not doing so is losing your non-English (or Japanese) speaking customer base. That non-E, non-J base is significant now (see here), and will only grow with time.

David’s presentation slides are below. Click on them for a full-size version. Japanese localized version is at the end.

Victor Alonso Lion and Diana Díaz Montón gave perhaps the most useful presentation of the day with their Localizing for the Global Spanish Speaking Community. The vast number of Spanish speakers in the world is daunting, made worse for marketers and localizers by the array of dialects. Victor and Diana did their best to synthesize all of the data. The information here comes from their presentation, an extended discussion I had with Victor during the SIG meeting two days later and some additional research. Any mistakes here are mine, not theirs.

The first thing to realize is that the Spanish speaking world is vast, located on two sides of the Atlantic. While the languages are all mutually intelligible, that does not mean that a resident of Buenos Aires will find it easy or pleasant to hear a Mexican or Madrid-based TV show. How many different versions of Spanish are there? Well, the answer becomes how finely you want to slice. Victor and Diana presented strikingly different sound bites of the same phrase from Spain, Argentina, Chile, Columbia and US/Mexico.

How different are these? Very. Simple terms in one country’s dialect might be a venomous curse in another. The polite word for “female” might mean “bitch” in another place. Further, basic terms in each language (pronouns, conjunctives, etc.) are different from one another. (I’d really appreciate if people could give examples in the comments.)

What is a game localizer to do in such a situation? The current solution has been to split the Spanish world into two parts: Castilian Spanish for Europe and Latin American Spanish targeted to the Western Hemisphere. While this works fine for Spain, it bulldozes the differences between the Western Hemisphere dialects.

Effectively, when we have our translators render Latin American Spanish, they create a neutral, understandable and inoffensive product which everyone can understand. David suggests Disney first created the concept of a single, neutral Latin American dialect to more effectively distribute their films.

The problem with neutral Spanish is that the language is not just neutral, but neutered. Imagine the scene from the movie Die Hard with Bruce Willis swinging down on the bad guys, bare feet bleeding, machine gun in hand, yelling, “Yippee kay yea, motherfuckers!” A neutral language might have to change that. “Comin’ to get you, boys?” That’s not memorable.

Is neutral Spanish bad? No, but it lacks emotional content. It would be hard to imagine a blockbuster — an entertainment product that strongly connects to people — using a neutral language. In fact, neutral language generally means lower perception of quality, both in the general population and, critically, in the local press. To maximize production values, different markets should be targeted.

-> Spain, with Castilian Spanish (the largest market)
-> US/Mexican Spanish, by far the richest market of Spanish speakers in Latin America
-> Argentina, with a population of 40 million and relatively wealthy
-> Chile, with a population of 17 million, but 50% more per capita wealth than their Argentine neighbors
-> Colombia, also ~40 million people but with half the GDP of Argentina

Another possibility, suggested by Victor, would be to mix the languages. For instance, a sports game could use a mix of well-known announcers from several countries: say Mexico, Argentina and Columbia. Spanish speakers would relatively easily understand (and forgive) speakers clearly from a different country, as well as feel attachment to the announcer from their country. It’s an elegant solution that limits the number of SKU’s a publisher must create.

Also, translation can be handled differently for the dialects to save time and money. An original translation can be done from English to Castilian Spanish, then a conversion from Castilian to the different dialects.

Another solution is to allow for amateur translations to be integrated into the final project. While this means potentially lower quality work in some cases, enthusiasts can do extremely good quality (note Wikipedia). Further, the amateurs working on the project then become evangelists for it, increasing market share.

Below are the slides from Victor and Diana’s presentation at the GDC Localization Summit in San Francisco, March 2010.

]]> http://www.apogeecommunications.com/blog/2010/03/22/gdc-2010-global-spanish-localization/feed/ 2 http://www.apogeecommunications.com/blog/2010/03/18/gdc_2010_localization_languages/ http://www.apogeecommunications.com/blog/2010/03/18/gdc_2010_localization_languages/#comments Thu, 18 Mar 2010 16:37:30 +0000 Edwin http://www.apogeecommunications.com/blog/?p=305 This post is about the surprising changes that came to light at GDC this year. The spectacular rise of Polish, and the disastrous fall of Spanish and Italian. The inevitable, short-sighted contraction of the language basket was indeed severe, but looks to be recovering.

2010 is a year of recovery for the Games Industry. The Great Recession hit with full force in 2009 and few were spared. Nintendo, home of the all-powerful Wii, was hit with a 40% drop in sales in June 2009. Less well-known, less powerful companies dropped even more.

The games industry reacted. THQ reduced their number of SKU’s (“stock keeping unit”, or the number of game variants released) from 1,200 in 2008 to 600 in 2009. Other companies reacted in similar ways.

Localization is an easy target in these times. Translation is seen as a large upfront cost and, given the gaming industry’s nascent international market, sales are dwarfed by the vast English language market (or substantial Japanese market), as seen here. Fully 82% of Take Two’s income is derived from English language SKU’s. EA cut back from its famous 10 language expansion in Europe, releasing one AAA title with only three full translations. Sometime in the future, when games will be a truly international market, these “savings” would never even be considered. That day is not here yet.

Bioware’s two AAA titles shipping this year had severely truncated language baskets. Spanish voiceovers were dropped from both titles, effectively removing them from the Spanish market. On the right is a slide taken from the Spanish language presentation at GDC; it tells the story eloquently — and the slide is from a pre-recession year. With 43% unemployment of their young people in 2009, Spain last year was perhaps a market best left alone. Italian voiceover was also dropped from the larger Bioware title, Dragon Age, leaving only French, German and … Polish.

Polish? Really? Yes, absolutely. Poland is the only European economy not to have entered a recession. This new powerhouse for Europe boasts the largest population of former communist nations with 40 million people. And the games market is clearly on the rise here. Poland is one of the few languages that certifies and awards well-done localizations (see the award on the left, hat tip to Rolf Klichewski).

Anecdotal evidence at GDC, and by other measures, things are very much improved now. Sales are going through, new releases are coming out, language expansion is being pursued again. 2010 is certainly shaping up to be an improved year over 2009. Possibly, with a little luck, 2010 may be a standout on its merits.

Some bits of data that weren’t included, or came out in the question-and-answer period:

Mass Effect 2 (450k words) was fully translated into four languages: French, German, Italian and Polish. There were text only translations into Spanish, Czech, Hungarian and Russian. There was an initial effort to localize fully into Russian that went so far as recruiting voiceover actors, but that part of the project was shelved due to market forces.

Dragon Age was fully translated into French, German and Polish. Text translations were done for Spanish, Italian, Czech, Hungarian and Russian. Given that there were between 900k and 1M words of text in this title, it’s understandable why the language list was shortened like this.

When asked about Asian languages, there was a “No comment” answer.

Each language had a relatively small group of translators in order to keep consistency throughout the work. Teams were composed of just three translators each.

The truly remarkable items in the presentation were the extensive Character Bible and string tracking facilities. If these were not in place from the very beginning, the project would have been quickly lost in a sea of changes as the games got developed. The other solid choice was when they decided that the text was locked down and ready for translation: one month after the English voiceover had been completed.

This was a very good, detailed nuts-and-bolts look at two projects that are much bigger than any non-MMO on the market today.

Below are snaps of most of the presentation’s slides. Click on any image to download the full-size photo. Please let me know in the comments if you have anything to add to this summary.

The biggest take-away from these slides is that localization is still in its early days. Companies are still relying too heavily on their home language markets. Between 20% and 30% of publisher income is from localized SKUs. Only relatively small parts of the world speak English or Japanese. Truly global companies will need to embrace the world more fully.

For an example, I’d look to Hollywood’s most recent blockbuster, Avatar. With some $700+ million in US revenues, the film has made of $2.5 billion worldwide. That is a close fit for the market graph at the top of this page. Avatar represents a truly international endeavor, and one which should be emulated by the major publishers in our industry.

These slides are free to use with attribution to Apogee Communications. If you want to see the underlying data, please email us. Click on any slide to see or download the full size image.

Sources of this data are:

Nintendo, Annual Report 2009, Kyoto Audit Corporation auditors. June 26, 2009.
Electronic Arts, Annual Report (Form 10-K) 2009. KPMG auditor. May 21, 2009.
Activision/Blizzard, Annual Report (Form 10-K) 2008. PricewaterhouseCoopers auditor. Feb. 27, 2009.
Konami, Annual Report (Form 20-F) 2009. KPMG auditor. July 31, 2009.
Square Enix, Annual Report 2009. Ernst & Young auditors. July 2009.
Take-Two Interactive, Annual Report (Form 10-K) 2009. Ernst & Young auditor. Dec. 18, 2009.
THQ, Inc., Annual Report (Form 10-K) 2009. Deloitte & Touche auditor. May 22, 2009.
Capcom, Annual Report 2009. PricewaterhouseCoopers auditor. August 5, 2009.
GDP Market Data from Economic Research Services, USDA, 2007.
Quebec GDP Data from Canada Economic Development for Quebec Regions, Nov. 2008

For the Gettysburg Address, it took 19 iterations and mangled it as much as you would expect.

Four score and seven years ago our fathers brought forth, upon this continent, a new nation, conceived in Liberty, and dedicated to the proposition that all men are created equal.

…becomes…

We are one of the two continents, seven years ago, for a new country, the same proposal, the ancestors of free time, many of us, to create a single-point one four.

Take this as a lesson — this technology is not quite ready yet.

The answer is indisputably, “Yes!”, though possibly not in the way you’d think.

First of all, the tools are essentially useless for production level text. The kind of gibberish they produce can glean basic information of what the text is saying, but any subtlety (and most grammar) is lost.

At Apogee Communications, we’ve had the unfortunate experience of reading some rookie contract translators using machine translation tools. They have the automatic translation done for them, then try to clean it up before sending it in. This work is always rejected before the client sees it, as it is never suitable no matter how much editing is done to it. Why? Because it proceeds from a word-by-word basis (dictionary lookup), rather than a concept-by-concept basis (“How would this whole concept be best rendered in my target language and culture?”).

We’ve seen this most often with Japanese to English translations (which are particularly excretable in machine language form), as well as some English to German translations (Why did they do that? These languages are close to one another anyway … How could you screw it up?). The results are never pleasing to the eye or ear and are always rejected.

So, how are the tools useful? Speaking as an active coordinator of Apogee’s many projects, it helps a great deal to verify parts of a translation. There are over 45 languages we deal with, and I certainly don’t know most of them. It helps immensely to have this automated tool to double check that the right translations are matched with the right source material.

Also, there is a review technique called reverse translation. Here you get one translator to handle your English->Chinese translation, for instance, then get a second one to translate that work back to English. It’s an expensive and time consuming process, but some clients insist on it.

Google Language Tools allow a project coordinator to get an reverse translation check done immediately. It’s not as good as a human resource, but can come up with some errors in spelling and basic concepts. Particularly in large projects, with lot’s of client changes in midstream, these tools are very useful for tracking strings to their proper places.

In the near future, I would not be surprised to see more fully functional machine tools for translation. Possibly even tools which would force companies like mine into more of a review function of automatically generated translations. But that day is not today. And I’ve been following this part of the translation industry since meeting the CEO of Dragon Systems some 20 years ago. Accurate machine translations are still a Holy Grail, though we are closer to it than we once were.

As discussed, the problem comes from several angles:

• Who is going to judge for the award? While we may have a few individuals competent to judge three, four or five languages, most big titles are now being translated into ten or more languages.
• If we choose to offer the award only for one title in one language (i.e., the German edition of Bioshock), we still need to compare different languages. Since this is always going to be a subjective judgment, who is going to decide?

The other option is to create a Wall of Shame, aka “All your base are belong to us.” This is a much easier process, since most fluent speakers can easily spot a bad translation. The Far East Economic Review had a favorite feature of signs written in bad English from India to the Philippines.

Either of these, the Award or the Wall of Shame, is subject to some political pressure. Any individual or agency will occasionally let loose with a ringer, which can be shown to all the world. If you look carefully enough, you are sure to find some which were penned by your worst enemy.