Let’s get this out of the way directly upfront - this is not clickbait. AI machine translation can be incorporated into your translation process immediately (as in, today). Depending on your process, you may be able to do it yourself, or you may need help, but it is 100% possible. Now that we have that out of the way, let’s dig in and see how we can start using AI machine translation immediately.
Can I leave it entirely up to the robots or do I need some checks and balances?
The answer to this question comes back to what you need to translate. If you’re translating emails from your colleague in Spain that’s sending you messages in Spanish - by all means jump right in. Product reviews? Website content (a lot of it)? - Let the robot do its thing. Do you need to translate marketing collateral or packaging for your medical device packaging? You probably will want to still incorporate a human linguist in the process. AI translation has come a long way, and it is fantastic, but it is still an automated process, and as I’m sure that anyone that has experimented with AI in any way knows, it is not perfect.
So the question to ask yourself is “Do I need these translations to be perfect?”. If you do, than you’ll still want to incorporate a human in the process. If not, AI might be just the ticket.
Another consideration before we jump in - Data Security
This is of course a really critical topic, and it also again relates directly to your content. If your content is publicly accessible and you have no concerns over data security or data privacy, then AI machine translation is a great answer. This also allows you to leverage free AI translation solutions, which have no guarantees around data security and data privacy (It’s free, what did you expect?).
If your content requires data security and data privacy then you’ll want to carefully review the legal agreements provided by any of the AI machine translation engines for both their free and commercial options. By the way, don’t just assume that the commercial options won’t use your data in a way that doesn’t align with your company’s data handling policies. Make sure you get into the small print.
I passed the robot check, now what?
You’ve determined where your content fits in the AI adoption spectrum, so now what? Let’s discuss the two ends of the spectrum, and how you can get started today (as promised):
Full AI Machine Translation
If your content is a good fit for full AI machine translation you can start immediately with several free options provided by any of the major providers (Google Cloud Translate, DeepL, AWS, etc.). Or you can pay for a commercial / business account which provides additional benefits like increased volume, and enhanced data security and privacy. These may be important considerations depending on what you’re translating.
We’re constantly reviewing the quality provided by the major engines and in general Google Cloud Translate and DeepL provide the best results, but there are definitely certain types of content and certain languages that you’ll want to route to one engine over another. Experimentation and testing is key. You may also want to consider using an LLM for translation (ChatGPT does a decent job), but as of the writing of this blog it’s relatively cost restrictive over a machine translation engine, and the quality doesn’t seem to be quite as solid as that provided by a translation specific technology.
AI-Assisted Human Translation
If you’ve decided that you would like to play it safe, but still gain the benefits provided by AI machine translation, you’ll want to consider incorporating a human translator into the process. The great news is that there are lots of different options for building out an AI-assisted human translation workflow. The most common workflow is an AI machine translation (let the AI do the heavy lifting) followed by an “edit” step performed by a professional human translator. The edit step can range from a simple clean-up to a full blown edit and review process. In our testing (again for certain languages and content types), we’re able to achieve a full human quality translation, while reducing cost anywhere in the range of 30%-50%.
While the AI translation with a human edit process is the most common, there are still other QA checks and balances that can be incorporated into the process to match your unique quality demands. It really comes down to your budget and timeline requirements. - Which may vary from project to project by the way.
What can I expect to gain by incorporating AI machine translation into my process?
So you’ve determined that you can incorporate AI machine translation into your translation process, and you’re deciding which approach will best match your quality, cost, and timing requirements. What can you realistically expect to gain once you’ve incorporated AI?
Cost Savings
You will see cost savings through the incorporation of AI machine translation. A main determiner in this is how much human involvement you still require to maintain the level of quality you need to deliver. Full raw AI translation could be as inexpensive as $0.01 / word. If the quality provided by raw AI translation is reliable enough for your content then you’re off to the races! I’m going to say this - and I’m not saying this to talk you out of raw AI translation, because I really do believe it has its place - please really consider whether raw, unchecked, AI output is safe for your use case. It is known to make mistakes, and in the wrong situation this can be costly, potentially wiping out any achieved gains. But, in the right situation, and potentially with a disclaimer, it can really shine from a cost-savings standpoint.
Time Savings
You will also most certainly save time! If timing is a critical priority for your content then AI could be a powerful solution. Raw AI translation is produced instantly. An AI-assisted translation process is much faster than a traditional full human translation process.
Quality?
If quality output is the highest priority in your translation process then introducing AI machine translation is not the answer. Now before any AI advocates come after me for that statement, I want to be explicit about my meaning. Adding AI will not improve the quality of your translation. It will definitively reduce cost and timing, but it will not definitively improve the quality of your translation. That doesn’t mean that AI produces low quality translation, it just means that it is not currently as good as pure human translation. One of the key technologies being worked on right now is the application of AI to grade and improve translation. Once that technology is in place we will likely see another big jump in the quality of the translation produced by AI, but until then you’ll just want to carefully consider what level of quality is the right one to match your specific needs and then go from there.
Conclusion
AI translation solutions abound and the entry to use is low, both in terms of cost and application. Once you finish reading this blog entry you could go to Google Translate and translate your content into multiple languages and send it out to print. As promised!
The answer for you and your content may be a bit more nuanced. Maybe you can use AI for part of it, and a more traditional process for other parts. Maybe you split it up by language.
If you have questions, would like some help or advice, or just more information, please feel free to reach out to us directly. We are always interested in hearing about how organizations are getting their translations done and finding out if we can help improve their current process!