Free to Choose: Unbabel introduces real-time artificial intelligence tool for selecting the best Machine Translation software

juli 7, 2023

The Dilemma with Machine Translation Selection

As Machine Translation (MT) has made remarkable strides over the years, the market is now more competitive than ever, with many machine translation providers and solutions available. The challenge, however, lies in pinpointing which MT model or solution can generate the best translation across multiple channels — this selection is crucial to maximizing translation quality and minimizing errors.

Generic benchmarks provide some level of assistance when comparing different MTs, but they have limitations when it comes to evaluating choices based on specific factors such as the domain, language pairs, and a business’s unique translation data. Moreover, running advanced offline tests on small content samples doesn’t yield reliable data, as the results don’t effectively carry over to real-world business use cases.

In a landscape where MT models are equally competitive,
making an informed decision is like finding a needle in a haystack.

Introducing Unbabel’s Instant AI-Driven Intelligent MT Selection

Unbabel tackles this challenge head-on with its groundbreaking Instant AI-driven Intelligent MT selection capability. Powered by Unbabel’s Quality Estimation (QE) technology, this feature empowers users to translate using multiple MT models available on the Unbabel Platform. They can also check the quality of each model and see the common mistakes they make.

The tool is launched with several third-party models (Deepl, Google, Microsoft, ChatGPT), as well as Unbabel domain models when appropriate, and more models will be added.

Aside from the document level comparison, QE also allows us to intelligently choose the best translation for each segment. This eliminates the need to stick with a single model; instead, users can select a model which picks the best translation per segment from a set of providers.

Unbabel’s MT selection technology is a game-changer, enabling users to automatically receive the optimal MT-generated translation for any source text, substantially cutting down translation errors. This, in turn, streamlines any human post-editing phases, making them faster and more cost-effective.

In an ever-evolving global market, Unbabel is set to reshape the translation landscape, once again, by offering unmatched quality, efficiency, and, above all, empowering customers with the freedom to make, objectively, the best choice for their businesses.

Ready to find the best quality MT that’s right for you? Try our MT comparison tool now.

About the Author

Profile Photo of João Graça
João Graça

João Graça is medeoprichter, Chief Technology Officer en rekenkundig genie achter Unbabel. De in Portugal geboren João studeerde computerwetenschappen op doctoraatsniveau aan een van de meest gerespecteerde technische universiteiten van Lissabon, het Instituto Superior Técnico de Lisboa. Tijdens zijn studie publiceerde hij een aantal goed ontvangen artikelen over machinaal leren, computationeel onderzoek en computationele linguïstiek - die allemaal de basis vormen van Unbabel's automatische vertaling engine. Na zijn afstuderen werkte João bij INESC-ID, waar hij onderzoek deed naar natuurlijke taalverwerking (NLP). Vervolgens deed hij zijn postdoc in NLP op University of Pennsylvania. João kreeg een Marie Curie, Welcome II Scholarship (2011), die hij afwees ten gunste van het ondernemerschap. Hij werkte samen met de huidige CEO van Unbabel, Vasco Pedro, aan de ontwikkeling van taalleeralgoritmen en instrumenten voor machinaal leren, en bekleedde verschillende functies als onderzoekswetenschapper voordat hij Unbabel mede oprichtte in 2013.

DeutschFrançaisNLdanskSvenskaEnglish