MODELA, the first Basque-Spanish Neural Machine Translator, is already available

The tool utilizes last-gen intelligent technology and has been developed thanks to the cooperation between four Basque Technology Centers, UPV/EHU and MondragonLingua

Starting today, the society of the Basque country can benefit from the best machine translation engine between Basque and Spanish: MODELA (www.modela.eus). The system has been developed using state-of-the-art neural network artificial intelligence technology, the most commonly used system nowadays: Neural Machine Translation (NMT). “We have achieved to apply this advanced tool, which used for translation systems between more widely spoken languages, on Basque documents as well”, said Itziar Cortes and Igor Ellakuria, heads of translation technologies and MODELA at Elhuyar and MondragonLingua.

Significant progress is being made in machine translation worldwide, but it still poses a significant technological challenge. In this regard, the Basque language has managed to enter the field of multilingualism and MODELA has represented a significant leap in quality. In the machine translation engines developed in the past, the quality of the results was limited due to the varied morphology and other specific characteristics of the language; however, neural network-based systems have managed to break through that barrier and the language tools for Basque have reached the same level as those for other languages.

Neural network machine translation engines provide an incredible level of fluency compared to the ones used in the past. “MODELA yields translations with a quality never seen before. For the first time, it has opened the doors for the practical use of an machine translation engine for the Basque language”, the heads of MODELA highlighted.

An opportunity to popularize the Basque language among Basque companies

According to a market study conducted by Lagune, the translation industry market of the CAPV (which includes both the private and public sector) amounted to a total of around 40 million euros in 2017 and a total of 500 million words were translated. Therefore, it can be concluded that MODELA will greatly boost technological and professional development, as it can, for instance, increase the productivity of translation professionals. However, it must be highlighted that MODELA is not meant to replace professional translators, but to create more content in Basque.

“It will be highly useful for people and organizations that want to use Basque as their language. Thanks to the quality of the translations, it will be possible to use Basque for day-to-day communication (internal memos, blog posts, etc.)”, according to the managers at Elhuyar and MondragonLingua. In other words, MODELA will promote the use of the Basque language in written communication and its standardization and companies will have a high-quality tool for anyone who cannot speak the language. This way, internal communication within the organizations that use it will be ensured and everyone will be able to communicate in their own language.

On the other hand, this high-quality Basque machine translation engine will be useful for the development of linguistic technologies and the creation of new applications, such as virtual assistants, text analyzers, etc., which will have a positive impact on industries such as healthcare, public administration, advanced manufacture and audiovisual, among others.

The result of the cooperation between four Basque technology centers, the UPV/EHU and a translation company.

The MODELA translation engine has been developed thanks to the cooperation between different agents of the Basque Network of Science, Technology and Innovation and it has been carried out with the aid provided by the Basque Government and SPRI through the ELKARTEK research program. The project has brought the most important Basque entities devoted to the research and development of linguistic technologies together. Moreover, the consortium has been in charge of all the steps of the process, from research to product marketing, with the participation of different parties.

On the one hand, three organizations that are at the forefront of translation technologies (the Vicomtech research center, Fundación Elhuyar and the Ixa group of the UPV/EHU) have taken the lead of the main technological aspects of the project and Ametzagaiña has been in charge of the general management of the project. On the other hand, ISEA, in cooperation with MondragonLingua and Eleka Ingeniaritza Linguistikoa (technology branch of Fundación Elhuyar), has been in charge of validating and transferring the results of the project to the market. It is important to highlight the significant contributions from MondragonLingua, Elhuyar and EiTB in the gathering of linguistic data and contents, which is a fundamental aspect in the development of machine translation systems. Lastly, MondragonLingua and Elhuyar have joined forces to increase the scope and accelerate the marketing of this machine translation engine.

More information and web service here.

MACHINE TRANSLATION

TERMINOLOGY MANAGEMENT

DOCUMENT TRANSLATION