Localization
Back to LAB main
Should the Games Industry Engage with Machine Translation?
The use of Machine Translation (MT) is becoming more widespread across every industry that wants to localize its content to different regions. The benefit is clear: the larger the audience, the larger the profits. However, the video game industry seems to be a bit late to the party when it comes to MT adoption for game localization.
So, what are the factors driving MT usage, and why should the games industry engage with MT?
According to a recent article, the global video games industry will gross $256.97 billion by 2025. Game development is no longer solely the province of AAA companies—now, more and more small indie studios are producing games. There is thus more content to translate than ever before. This content is not confined to spoken and on-screen dialogue and text, but also occurs in tutorials, documentation, support chats and emails, videos, blog articles, technical specs, etc. Every piece of text and audio in and around a game is available for translation. This is a huge and potentially overwhelming amount of work for translators.
Proper post-edited machine translation can ease the burden, saving time and effort. Simply put: more languages means a wider range of audiences, meaning that a diversely translated game can reach brand-new markets. Because games come from all across the globe these days, it makes more sense for a producer to assess the needs of the game, run a feasibility study and appropriately deploy MTPE (Machine Translation Post-Editing) where possible.
Game development is a long, painstaking process. Localization can take just as long if the internationalization process hasn’t been considered from the beginning. The smart developer will account for this time and use MTPE wisely to shorten the time to market. At this stage, there are certain kinds of content and language pairs that MT handles better than others, so this can be a factor too.
Native linguists can assist with training the MT engine to improve its fluency by identifying key terms and phrases and ensuring the engine gets it right. Compared with traditional human translations, the combination of MT usage and strong post-editing can increase productivity from 15 to 40%, vastly improving release windows if done right.
There is always a balance between quality and cost saving, and a large part of that is assessing which MT engine is right for your game. Being engine agnostic and conducting feasibility studies provide the necessary insight to make these decisions, but it’s clear that usage of MT offers great improvements in cost savings. It works particularly well for languages for which MT already has a proven history of success: a well-researched lexicon and the kind of content type that translates quickly and easily. Not every project can benefit from MT post-editing, but it’s always worth the initial analysis.
There is in-game textual content that requires localization, like
character dialogue and user interface items. But gaming companies have other
types of content to translate that is suited to the MTPE workflow. User-facing
content like support tickets, chats with players, and cross-country communities
where one discusses how to play their favorite games are all fodder for
translation. At PTW, MT has been
tested successfully and deployed in Player Support to improve coverage and the efficiency of
responses, when questions are asked in languages the expert agents do not
speak.
Having mentioned all the above, MT is not a one-size-fits-all solution for every scenario. In-game content can be challenging for MT, as it may contain idiomatic dialogue (featuring slang the MT engine wouldn’t understand), buttons with one single contextless word (that might confuse the MT engine into replacing with the wrong word), and humor (which can rely on local-specific contexts) or fantasy (that often contains new words or phrases specific only to itself, as opposed to real-world terms or phrases). Thus, full post-editing is needed, or alternatively, PTW may advise using the traditional localization workflows and look for customization options. However, an MT engine is still useful for other scenarios, such as help articles, blog posts, legal requirements, technical specs, etc. The cost and time savings in these cases make themselves evident.
MT is most useful for literal, word-for-word translations. This was well-known in the past and it is the reason why other industries like science and technology, engineering, construction, automotive services, etc., benefitted from MT in the early stages of its development.
Traditional translation was seen as adequate for most video games for decades. But social media has proven that players care about properly translated texts, and these days the threat of negative word-of-mouth can be a powerful motivator for businesses who care about their reputation.
With more recent developments in Neural Machine Translation (NMT)—like probabilistic models and artificial intelligence—MT engines are more adaptable and fluent than their predecessors. Some engines may even take pictures and visual aids as context to disambiguate a word or a sentence, thus improving the MT output quality and reducing post-editing effort.
The more the games industry involves MT in its processes, the stronger the tool becomes. At PTW we have seen productivity gains go from 10 to 20%, and even up to 50% for some language pairs and types of content. Our MT consulting service assesses the specific project needs to recommend one engine or the other, or can even customize one from scratch. Done effectively, you can achieve high-quality results using MT.
The games industry has always been at the forefront of adopting new technology, and MT should be no different. High volumes of content can be localized far more quickly using MT than the traditional method, and it can be a cost-effective solution for reaching more players than ever before.