This article was originally published on InLingo.*
Natalia Gladkaya has always loved linguistics, but she never expected to one day merge her career with video games — it seemed like a hobby that would never generate income. Everything changed nine years ago. Her hobby became her job, and now she leads the localization office for an international game developer.
We spoke with Natalia and learned how the localization office at Plarium works, how to work with Games as a Service, and why translators should already be thinking about retooling for post-editing.
— Did you play video games as a child, Natalia?
— Like a typical kid of the nineties, I started with more mass pop culture — translations of literature and films. I was interested in languages, so when I read serious works of fiction in translation I couldn’t help but think: “If only I could read this in the original and understand the work that goes into conveying the same ideas in other languages.”
My interest in games started with console games — first we hunted ducks (editor’s note — reference to the game Duck Hunt), then we moved to FIFA and Heroes of Might and Magic. I think that’s a pretty average trajectory for the beginning of the 2000’s. Video games were also translated from the beginning. The games had varying quality and reputations because of their localizations. For example, GTA: San Andreas became a meme because of its Russian translation done by pirates.
I never planned to merge my profession with game localization — I was more interested in technical translation and fiction. Games were just a hobby, but then they intersected with career opportunities and I thought: “Why not?”
— How did you first get into the games industry as a professional?
— I joined Plarium in 2011. Back then, the company wasn’t anything like it is now. It was more like a fast-growing start-up, and that worked to my advantage. I had a linguistics background, but I had never worked with game applications and had a very poor understanding of the development process. The company’s starting positions allowed me to get on the team as a translator and proofreader.
— Are there any challenges linguists encounter in the gaming industry that they wouldn’t in other fields?
— Absolutely. The game environment is quite specific and has lots of restrictions. For example, there are strict requirements for a text’s readability and the number of characters. The game interface isn’t big, so you need to convey as much information as possible, but say it in an accessible and efficient way. Every industry has standards, but in this case they change depending on the type of game, its genre, its area, and its target audience. It took me a long time to get used to that.
Any translation should take into account its target audience, but that is particularly important in the game environment. It’s one thing to just create content, but another thing entirely to make sure it suits its target audience and doesn’t create dissonance — that is, it doesn’t diverge from the game experience and the player’s expectations.
Production requirements are always updating. When I started working, out of inexperience I thought we just launch a game on Facebook and then work off that framework, since all the processes are already established. That was partially true, but not entirely: product features vary greatly, and our approach to each project depends a lot on them.
Nords: Heroes of the North game that Plarium released on Facebook.
— When does the localization team enter the development process?
— We did a lot of experimenting with operating formats and tried to understand when the right moment was to start the process. It’s best to integrate localization with development as early as possible. It’s pointless to start too early — that’s like painting walls that aren’t even built yet. A product can change a lot during development, so we have to find the sweet spot and start at the right moment. It can save a lot of time and energy.
We spent four years forming the team’s operating standards, and we realized that we needed to start working with content at least during layout creation. At that point we understand approximately how a game will look and what kind of experience we want to give the player. We know what to let them figure out on their own, and where to show our hand.
— Is the localization office included in the decision-making stage, when it’s still too early to start working but a game’s primary features are determined?
— We are very much tied to marketing and are brought in when the decision is made about which markets a product will target. We establish a list of languages, do additional research, and make a decision. However, the basic list of languages changes frequently during the discussion. Some markets can wait, but we try to access others right away to capture a new audience.
— Do you have any criteria for assessing the prospects of this or that market?
— Of course, we have such criteria. We look at general demographics — a country’s population. If the product is designed for mobile platforms, we find out what percentage of the population uses cell phones, which device models are most popular, average screen time, and which projects have been the most successful there. We basically try to gather as much publicly available statistical data as possible.
Raid: Shadow Legends, one of the most popular Plarium projects, is localized into 13 languages including EFIGS, Asian languages, Portuguese, Turkish, Russian, and Ukrainian.
We also try to accent cultural and linguistic features — for instance, how anglicized the market is. In the Netherlands, for example, everyone speaks the native language and uses digital products in Dutch, but their game market is practically unlocalized. It’s the same story with foreign shows — they watch them in the original language or sometimes with subtitles, because most people there have a good command of English. So, that begs the question: is it worth localizing a product there in Dutch?
Every business decides that for itself, but we consider this a situation where translation to the native language is not necessary. On the other hand, in Portugal, English is not really used commonly, so audience won’t be able to fully enjoy an unlocalized product. From this viewpoint, having Portuguese in our games is more essential than Dutch.
— Does localizing for new markets allow you to attract not only new players, but a wealthier audience?
— We don’t have any data suggesting that a well-off audience was attracted specifically because of localization. We can assume that it indirectly helps increase the audience and bring users to a game who are ready to pay for it, but we haven’t been able to track a pattern or graph it. It’s hard to be exact here.
— How does Plarium choose translators: do you work with an in-house team, vendors, or freelancers?
— We use a hybrid system — each language and market requires an individualized approach. Most often we use freelancers who are selected based on certain criteria including whether they have an interest in games and possess industry knowledge. That’s not always the deciding factor, but it’s a major plus for bringing freelancers on board in a project. We try to integrate freelancers in the development process as much as possible. Games as a Service (GaaS) are an ever-changing product, so it’s very important for our specialists to have a certain level of understanding and flexibility.
However, not all markets and languages work that way — in some cases we look for vendors. Most often they are small companies that specialize either only in localization or in a wider range of services. Whatever the case, it’s important for us to integrate people in the process, whether they’re freelancers, small studios, or large vendors.
Specialists need to know exactly what is required of them. We facilitate that by regularly updating the style guides and instructions we send to translators. If you’re creating a GaaS that is constantly developing and gaining new mechanics, don’t forget to update the guides.
— And who handles style guides and glossaries — an in-house team or proven freelancers?
— At the beginning we compose a basic style guide internally with general localization recommendations and requirements. Then we deploy specialists in certain languages to review the style guide, make changes, and offer recommendations. For example, they say: “This joke won’t work in Germany — German humor is very different from the humor of other central European countries. Let’s come up with something else.”
— How do you select project managers? Is each of them attached to a specific project or is it a team of super experts that work on everything at once?
— At first we tried to train universal localization managers, but we found that, because of the large number of projects and their long lifespans, it wasn’t very easy to transfer managers from one project to another. The processes are similar, but they’re still all different in some way, so we try to operate so that every localization manager is responsible for their own project. In some cases a manager might have several projects — it all depends on their scale. Sometimes we know that a certain specialist handles military strategy games best, so we choose them for such projects. We assemble our translator team in the same way.
Again, we come back to the style guides and streamlining work with linguists, because that way we can use the same translators to work on most games. It would be strange if the same specialists receive entirely different requests from managers, so we try to find a universal formula to apply to different projects. However, that process can change depending on the specifics of each game.
— If you plan to release a game on the German market, but a localization manager doesn’t speak that language, would they be assigned to that project?
— Speaking the language we’re localizing into helps, but speaking English and understanding project management are the priorities here. If we know we’re launching a game on the German market but we don’t have a localization manager who speaks German, we will not change the current manager. In the end, we have German translators.
At the same time, it’s important to understand the features of the market you’re working with. We compose style guides in universal English, but we still pay attention to the wording and periodically change them depending on the mentality or business culture of a certain country. Our amazing team of Spanish translators might go above and beyond, and then watch a video on YouTube and write to us: “Check out how this dude balanced your characters from the player’s perspective — your game designers should take a look.”
Most of the time having a dynamic and warm relationship with the team is a pleasant bonus. Business relationships take years to build, and it doesn’t matter whether the team is made up of vendors or freelancers. We try to establish long-term relationships with people, which helps us understand one another on future projects, because we get used to each other. Business processes harmonize, and it gets easier to find a common language with time.
— The defining feature of the GaaS model is the updates never end. How do you make sure the localization process works smoothly?
— This process requires constant work, because the demand for GaaS is relatively high, and competition is growing — whoever captures the audience first and holds its attention wins. The trend has been towards speeding up the process, and increasing its effectiveness and optimization. There needs to be a balance between making something cool and refined but not delaying the product.
If you are able to maintain around-the-clock localization support, that’s great, but you have to understand that that requires a whole lot of resources. It’s always desirable to speed up quality translation, but there is a minimum time requirement for translation and LQA. Sometimes you want to skip LQA, especially for smaller updates. That is always very tempting in GaaS, since you can fix mistakes in production, but it doesn’t always work in our favor.
Most bugs really can be fixed without waiting for the next patch, but sometimes changes after release can backfire. For example, if we introduced a character who was not very well localized, but players are already used to their name. There might be a better option, but the character is already active, and changing anything might confuse players. This raises the question: is it worth changing the localization and retraining players or should you just let the name become a meme in your gaming community and move on?
In general, the less time there is between releases, the more time you need to spend, however strange it may sound, on contextual support — the localization kit, for instance. The more information we can provide linguists at the beginning, the faster they will be able to produce high-quality localization. That way, if we didn’t have time to test a certain language, there is less of a chance that something goes wrong. Good lockits and style guides save a lot of time during LQA.
— Let’s say we have a project that is several years old with a set team working on it — a translator, proofreader, and tester. What happens if the translator suddenly gets sick and can’t work?
— It’s best to be in touch with all the specialists, even those working remotely, in real time. Then we decide: taking into account the current situation, can we wait or do we need to act right now?
We can engage a translator from another project or ask the proofreader for help. If the LQA tester speaks the source language, they can try to produce a sample version without final approval. I know some companies use machine translation, roll it out in the game, and only then use a native speaker to edit it. We’ve never used that strategy ourselves, but if your engine is well trained and produces an adequate version, it could work. That is, as long as it isn’t a Google Translate translation, which, although it’s improving, still can’t produce appropriate quality for a game release.
As for machine translation, we use it for pseudolocalization in the initial stages of project development. We generate sample content to make sure fonts and various technical elements are displaying correctly. We stay in touch with our colleagues in the industry, and some would say that you can save a lot on localization services if you use machine translation correctly, followed by editing. Our team isn’t there yet — it’s our impression that automatic translations create a lot of additional work. I think engines aren’t ready to replace real people yet.
— Do you think what everyone is afraid of will happen in five to ten years — that machines will replace people and translation will turn into post-editing?
— I can’t say about five years, but I would place a small bet on that happening in the next decade. There’s already a lot of promising technology that is used to train engines in real time — you edit a segment and the engine uses that data entry in the future, not by auto-filling or concordance, but by using semantic analysis, if I can call it that. The first steps have already been taken in that direction. That doesn’t mean translators will be out of a job and sit at home while neural networks do everything. It’s still too early to think about — everything depends on the type of text, so human editing will still be necessary.
— Which CAT tool do you use at Plarium?
— We use memSource, which was made by Czech developers. We chose it because it’s the simplest and most intuitive of all the CAT tools — it’s the easiest one to figure out. If you compare Trados and Memsource, for example, Trados has a lot more functions, but it requires a lot of per-linguist and per-project customization.
This is what the Memsource interface looks like. Picture Source: Memsource.
We have a large, spread-out team, so we use a simpler option. It’s especially great that memSource seamlessly works with any other CAT tool. I mean, our freelancers, for example, can work in another tool that they are more familiar with and convert it to memSource with the click of a button.
It was also important to us that our CAT tool connected with our own product, allowing us to load text into a game. Memsource can do that.
— What project management tools do you use?
— We use Jira and everything that connects to it — Gantt and Kanban. Localization, management, and development are really closely related in our case, so all of our boards are integrated in one main project.
Jira is a universal tracker for us. It’s convenient for entering statistics, project planning, and monitoring task completion. We also use good old Google Sheets, no one has come up with a better or clearer option.
— How do you plan out your workday?
— In the morning I handle emails and planning. Before quarantine, we used the Stand Up format where everyone gathered in a circle and discussed projects or shared news. The afternoon is when we start on current tasks and the more hardcore process. Around 7 PM someone always writes: “Guys, we have this really urgent task here” or “I have a great idea.” That’s usually the end of the day.
Everything also depends on the development cycle: if we’re at the very beginning of it, everything is more orderly, but as we approach release, the process gets more chaotic. I feel like that’s true in any area.
— If there was only one Plarium game left and it wasn’t Raid, what would you play?
— I’d say Mech Arena, despite the fact that I’m not good at PvP shooters, especially as a mobile app. It’s a cool, captivating, and very dynamic project. Team battles of giant robots in an arena — what could be better? If not Raid, then Mech Arena.
— What professional media do you read to stay up to date on the industry? It can be game dev or linguistics.
— Game dev is pretty easy — I read the most popular English publications: Game Informer, IGN, Polygon. I like the Russian publication DTF, but that’s a mix of films, games, and pop culture in general. But it has customizable sections. I check out DTF pretty frequently for feedback on Russian localizations. It’s helpful to read it because it changes your perception of a project and you see how end users view it. It’s a pretty interesting life hack to improve service quality.
As for linguistics, it’s harder for me to name individual resources and portals because they are most often broken up into categories. Some look at localization topics through the lens of marketing and others through community management. LinkedIn has communities of localization managers from various companies.
They share experience and help create an atmosphere for networking — you can always find someone else who has encountered the same problems. Besides that, you can ask them about a specific market, so you know, at least in general terms, what to expect.