“Google Translate” has become synonymous with automated translation, but there are many forms of computer-assisted translation. With the wealth of translation technologies available today, many organizations are exploring options to figure out which tool works best for them, and when to bring in a human editor (our take – always!).
What is Computer Assisted Translation?
In the translation industry, computer-assisted translation refers to a set of tools designed to empower human translators to do their work efficiently.
Computer Assisted Translation (CAT) tools like Trados (Alboum’s primary tool), memoQ, and Wordfast enable greater consistency and support machine translation processes.
Outside of the translation industry, computer assisted translation is a general term that is often used to refer to the different translation technologies available. Our goal with this post is to help you make sense of the options, so we’ll use the term in a general sense.
Which tool you use matters.
Computer assistant translation can improve speed and save money, but there is a significant difference between something fully-automated and a process that pairs technology with humans to refine the output.
We believe an editor is essential to ensure quality, consistency, completeness, and cultural accuracy of any translation.
Machine Translation (MT)
While not technically an “AI” solution, machine translation automatically matches words with their foreign language equivalents. Google Translate is an example of machine translation.
Most free machine translation tools are untrained models that provide literal word-for-word translations. Free versions of tools like Google or DeepL can give you a good sense for the overall meaning of a text, but shouldn’t be considered publication quality. This is especially true for character-based languages or those less commonly spoken.
Paid versions of those tools or others, like AWS or Language Weaver, offer a better-trained solution along with security and privacy features. However, unless integrated with a translation memory tool or process to support a human-in-the-loop solution, paid tools may not bring the consistency and clarity many organizations desire. Alboum uses a combination of translation technologies in our MTPE solution.
Pros of Machine Translation include:
- Great for getting a sense for overall meaning of a text
- Accurate for common terms, widely-spoken languages
- Speed and affordability
Cons of Machine Translation include:
- Low quality when translating complex content, homonyms, idiomatic expressions, or text with cultural nuance
- Final output can be inconsistent, particularly when the same term is used in different ways throughout a document
- Errors can be significant with character-based and right-to-left languages
- Low quality for languages of lesser diffusion (those less widely spoken)
Generative AI Translation
Generative AI Translation takes machine translation multiple steps further.
Generative AI translation leverages large language models, vast amounts of data, and adds in contextual clues – creating new foreign language, human-like content.
Examples of generative AI translation tools include ChatGPT, Lokalise, and DeepSeek.
Pros of Generative AI translation include:
- Constantly learning so accuracy improves with each additional translation
- Ability to process slang, colloquialisms, and cultural nuance
- Automated glossary and machine translation memory updating
- More natural sentence flow
Cons of Generative AI translation include:
- Tools struggle with exceptions, even when rules are well-written
- Unreliable when accuracy is critical – such as in legal or medical documents
- As learning and data storage is integral to how the tool works, there are significant privacy risks
AI Interpretation
In addition to computer assisted written translations, there are a wealth of real-time AI interpretation tools including:
- Apple’s AI translation-enabled AirPods
- Meta’s Glasses
- AI voice translation on Zoom and Google Meet
AI Interpretation brings together multiple AI solutions to support the output, including:
- AI transcription
- AI translation
- AI recording
These tools are newer, so there’s work to be done on quality. They’re fun to play with for interpersonal conversations, but not quite ready for prime time when accuracy truly matters.
Beyond the quality issues, there are also privacy concerns. You should expect that any conversations processed through the free AI interpretation tools will become part of the AI’s training data, creating a permanent record outside of your control. The EU has gone so far as to block Apple’s live translation feature, likely because they are not compliant with GDPR regulations.
The Bottom Line on Computer Assisted Translation
As AI and apps in general continue to evolve, multilingual features will as well. They are fun to play with and great for quick, interpersonal needs.
But, for government agencies, schools, healthcare providers, and more, choosing the right tool paired with an editor makes a world of difference between providing real language access and experiencing embarrassment over the wrong word choice or legal exposure over privacy violations.
Learn more about our computer-assisted translation process, or contact us today to request a free quote.
