Rapid Translation of Open-Ended Responses Through Machine Translation and Human Post-Editing
It is increasingly evident that speed and agility have become hallmarks of the insights industry. THG Fluently is a longstanding provider to most of today’s leading global market research companies, and a specialist supplier to the sector for almost twenty years. Our alignment with the industry has shaped our service provision and our permanent search for innovation.
We know that finding insights faster creates commercial advantage. This is why we are always seeking to add new and improved tools to our service portfolio that will help the sector win in both reduced turnaround time and cost efficiencies in an increasingly competitive environment.
Savanta – a key player in the world of B2C and B2B market research – had a particular project for a major client in the gaming industry. It involved verbatim translation and subsequent coding amounting to almost 57,000 words across seven languages (German, French, Italian, Spanish, Dutch, Polish and Romanian).
Savanta’s in-house coder only supported English responses, so THG Fluently were brought in to support the multilingual elements of this research.
The client needed a quicker turnaround time, with a shorter delivery deadline compared to the typical eight-day turnaround time for verbatim translations.
Accuracy was also a key request, with the client seeking to gain insight from the specifics of responses as opposed to gist translations that would otherwise give them a less in-depth insight into the nature of the various responses.
THG Fluently assessed the requirements of the project and, in light of the need for both accuracy and a quick turnaround time, determined that a light touch machine translation (MT) solution would be suitable.
This approach combines the speed of MT and the enhanced accuracy of the MT output thanks to an additional human-in-the-loop stage, in which a linguist reviews the machine-translated content to ensure mistranslations are minimised.
A common challenge of machine-translated verbatim and open-ended responses is the varying register of the language used. On occasion, survey participants use slang and idioms that can be mistranslated or taken out of context by an MT engine. However, the human post-editing step ensures these challenges are addressed appropriately and that the content is of the desired quality and accuracy.
In Savanta’s case, the light touch MT post-editing solution enabled the delivery of the translated files within three business days, including all internal QA steps. This is a 63% reduction in turnaround time versus a traditional translation workflow.
“THG Fluently has been our preferred translation vendor for many years, and we've never had to consider changing this — they are always responsive and accommodating, even with last-minute requests and changes. Always offering good solutions in any difficult situation, making valuable recommendations, they're therefore our most reliable partner with any language-related request. Their combined approach of machine translation together with human translator post-editing helped us massively on a large multi-country project where the timings were the key. This made a very challenging turnaround doable and helped to keep our client happy.”
(Associate Director, Savanta)