Automated QA
As well as the engines themselves, technology exists that carries out quality checks on machine output in an automated fashion, such as TAUS’ QA tool. It works by finding linguistic errors such as misspellings, terminology errors, and formatting errors, among other things. What automated QA can’t do, however, is replace a human reviewer when it comes to checking output for accuracy of meaning or cultural appropriateness. Nonetheless, it can be a good starting point for flagging potential issues that should then be checked by a human.
Machine + Human
In our opinion, the best solution for the use of machine translation is in combination with a human linguist. The proportion of machine output versus human output for a single text depends on the need and also on how much of the content has been translated before. However, by re-using existing translated content where possible and generating machine output for content lacking previous translations, which is then edited by a human linguist, produces a final publication in which the confidence level of accuracy is high compared with pure MT output.
Choosing the right type of AI-assisted translation service and even whether AI-assisted translation is the most accurate and cost-effective solution for you, will depend on a number of factors, the most important being the type of document you are translating and the size of the document. Creative content such as marketing prose is not often handled well by AI. Similarly, safety-critical content should not be trusted solely to a machine without human input. As such, we provide a range of machine translation services with the aim of catering for different requirements.
Get a Quote
