Do you remember the last time when people were NOT talking about machine translation (MT)? We don't. Wherever you go, there’s someone talking about MT. With few exceptions, it seems like the only major disruptors in our industry over the past few decades have been breakthroughs in language technology.
However, as we discussed in a previous Nimdzi report, the hype is not completely justified, at least not in any way that is clear and apparent. Nimdzi surveyed 33 localization buyers to ask them if they were currently using neural machine translation (NMT) in their organizations. The replies confirmed that NMT is not as widely adopted as one may be led to believe.
Source: Nimdzi Insights
In this survey, 77.4 percent of buyer-side localization managers responded that NMT is far from being fully adopted within their organizations. What’s more, 33 percent said NMT isn’t being used at all.
In the games industry, specifically, there have been a few attempts to include MT in the localization workflow. For example, Electronic Arts (EA) was an early adopter of this technology and has learned how to make the most of it, even in a creative industry. In an article published in the magazine MultiLingual titled “The future is here: Neural machine translation for games,” MT specialist Cristina Anselmi and localization veteran Inés Rubio shared their insightful experience implementing MT in the game localization workflow. In the article, the authors shared a table with the different areas and content types at EA where MT is applied.
Categorization of types of text at Electronic Arts. Source: MultiLingual
As we can see, most of the text categories where raw MT is implemented don’t have a checkmark in the category “Sentiment,” meaning it’s not content that has a direct impact on players’ emotional engagement. Game content and websites do have a direct impact on players’ experience. In these cases, EA uses post-edited machine translation, ensuring the final text is carefully reviewed by a human translator. And on top of that, they use this workflow only for titles with low ROI for specific languages or in low impact locales with low ROI.
EA has developed a very smart way to use MT to streamline their processes without having a negative impact on the quality of their players’ experience. This system shows that they are extremely mature in the way they understand their content and its impact on their users. But this also supports what we see as a clear mantra for the industry:
EA’s use case is just one example of how MT may be used for locales or products that are perhaps less important for a company at a given time, or even for content that doesn’t have a direct impact on the quality of experience of end users. MT certainly has its uses for customer support or documentation content that is not highly relevant for the actual gamer experience. But, when the revenue of a company is at stake, companies usually prefer to go with human translation, especially in creative industries such as media and game localization.
We recently introduced you to the two- (or five-) second rule, which is essentially the reaction or decision-making time a linguist should spend judging whether to post-edit a segment of machine translation (MT) output or to retranslate it. This rule of thumb aims to help increase the linguist’s productivity when working with MT.
If you’re a driver, you’ve probably heard of the two-second rule. Staying at least two seconds behind any vehicle is considered a rule of thumb for drivers wanting to maintain a safe following distance at any speed. The two seconds don’t represent safe stopping distance but rather safe reaction time.
Some machine translation providers are holding out hope for MT systems that adapt to document context. Could this development eliminate the need for custom MT engines? Will context-enabled MT help MT achieve human parity? Will we still need to customize a few years from now? Let’s discuss further.