TAUS and Intento, two companies specializing in language data and machine translation, have teamed up to form an international partnership focused on training more efficient and effective machine translation (MT) models. The two companies hope that their partnership will help improve the MT landscape, by combining TAUS’ specialization in language data services with Intento’s focus on MT training and evaluation.
“By combining our expertise and capacity to generate, collect and annotate language data with Intento’s MT evaluation and training expertise, we expect to offer an accelerated transformational value in the MT space,” said Jaap van der Meer, director of the Amsterdam-based TAUS. “Up to 90% of machine learning models do not make it into production. The key differentiator is the use of data that is fit for purpose and high-quality that make these models perform at an optimum.”
TAUS, which was launched in 2005, works to provide companies with language data services and has grown to offer one of the largest data repositories in the industry. The company is also known for their Human Language Project, which recruits a global team of translators to build a corpus of data for less commonly spoken languages in order to ultimately improve their accessibility online. Berkeley-based Intento works to develop artificial intelligence services (AI) for numerous companies, with a particular emphasis on MT. The company uses training data custom-tailored to clients’ needs to boost efficiency, and has also formed partnerships with numerous other LSPs.
“At Intento, we help global companies procure and deploy the best-fit AI services and we acknowledge that AI and ML performance depends on the quality of the training data,” said Konstantin Savenkov, chief executive officer of Intento. “By partnering with TAUS, we ensure that the highest quality of language data available fitting the purpose and project specifications will be used in all of our MT evaluation and training projects.”