Connecting Cultures and Expanding Communication

At 7C Lingo, we are always excited by new developments in translation and interpretation technology. We firmly believe that integrating initiatives such as machine translation into your pipeline will not only result in quicker project turnover, but will ease the burden of the language experts you work with. It seems like we are not alone in our assessment of this technology, though, as everyday, new innovations are debuting and pushing language services further and further. With so much happening so fast, we wanted to highlight three projects we expect to make waves very shortly.

Signs

AI and Language Services have had a somewhat combative relationship since the technology debuted, with many experts in translation and interpretation remaining sceptical on the efficacy of technology like machine translation. Still, many companies are choosing to work with these experts in order to push the entire industry forward, with one of the most promising efforts being Signs, the newly launched ASL instructional platforms.

Developed by Nvidia in partnership with Hello Monday, Signs is a cutting edge web platform designed to help teach American Sign Language. Powered by AI, Signs gives real time feedback to ASL learners, providing feedback and ensuring learners steadily increase their accuracy. This is paired with a massive database of ASL signs, demonstrated via video clips, that have all been validated by ASL experts.

While the software launched in February, Nvidia has already pledged to continue expanding the program. During a press conference, the company stated they aimed to include over 400,000 video clips showcasing around 1,000 ASL signs, ensuring that any user of Signs has access to a wide vocabulary. While AI powered ASL interpretation is probably years away, using AI to instruct the next generation of interpreters is breaking new ground, and will be exciting to monitor as it develops.

SMOL

A huge issue plaguing AI translation software is a lack of datasets available for certain languages. With many AI models being trained on data scraped from the internet, the overabundance of content in English has resulted in the model’s lacking accurate training data for many other languages. In order to rectify this, many companies are developing training datasets, including Google.

SMOL (Set of Maximal of Overall Leverage) is a data set constructed by professional translators designed explicitly to provide translation data for 115 low-resource languages. This release consists of two elements: SMOLSENT, a collection of 863 English Sentences that have been translated into 81 different languages, and SMOLDOC, which consists of 584 English documents of various topics that have been completely translated into 100 languages. This second set is extremely promising, as many AI software struggles with translating technical documents for legal and medical purposes.

Researchers have already seen an improvement in translations after these datasets were fed into the models, showing off how effective initiatives like this can be. In addition to its robustness, the dataset is also open-source and freely available via Hugging Face, allowing for anyone developing an AI translation software to use it for development purposes.

Vasco Translation Earbuds

Translation has been at the forefront of discussion for technological developments in language services, and for good reason. While programs like machine translation are constantly progressing forward, similar efforts for simultaneous interpretation have struggled due to the added complexity of things like tone of voice, regional pronunciations, and even background noise. Still, progress is being made, with Vasco being one of the companies at the forefront of the movement.

Already, Vasco has made waves in the industry with their series of Portable Translation Devices. These devices are similar in size to a cell phone and can translate 82 languages via voice, and 112 languages via photos. Always searching to provide the customer with a lower profile option, Vasco launched the Vasco Translator E1, a pair of earbuds advertised as being able to interpret up to 51 languages in real time. After installing the Vasco app on your phone, just select the language you wish to interpret and Vasco will feed directly into your ears.

This technology, though, is not perfect. Many users report having to speak slow and deliberately in order for the earbuds to function, and the steep price tag of $389 will definitely be a deterrent to many. Hopefully, though, the technology will continue to push forward and, in time, people will be able to hold a normal conversation using these earbuds.