As with everything in the AI world, the use of AI in translation is moving quickly. Just a few months ago, we wrote this primer on translation technology tools. Now, DeepL, Amazon Translate, Google Translate, and many other commonly used machine translation tools have integrated AI into their models. While the promise of better quality remains, current results are inconsistent.
Given the constant state of evolution of machine translation tools, inconsistency is expected and able to be both identified and overcome. That’s why post editing is a must. Unedited machine translation, particularly AI-enabled machined translation, is subject to:
- Inconsistent translation of the same word within the same document
- Deletion or distortion of key terms
- Hallucinations
- Privacy risk (when done in free tools)
The Language Access for All Act currently making its way through Congress will codify this. The Act requires human involvement in all translations produced by government agencies and those receiving Federal funds.
Alboum continues to recommend machine translation with post editing for many situations. It is a great option to reduce cost and speed production time. But as the adage goes, garbage in, garbage out. Organizations can play an active role in improving the quality of translation projects (regardless of the approach) by:
- Proofreading for punctuation, grammar, typos, and any other errors
- Editing text to streamline repetition and ambiguity
- Creating glossaries of key terms
- Sharing style guides and existing acceptable translations or translation memories
These actions empower linguists to deliver the best translation within the specified timeline.
Contact us to learn more about how Alboum’s human-in-the-loop translation process can help your organization.
