AppTek Announces Neural Machine Translation System
McLean, VA, USA Jan 22, 2019 -- AppTek, a provider in Artificial Intelligence, Machine Learning, Automatic Speech Recognition and Machine Translation, announced the availability of its multilingual neural machine translation (NMT) system that performs translation of 12 Slavic languages into English. Customers can install the system on-premise or use it via a cloud-based Application Programming Interface (API).
The 12 languages all come from the same Slavic family. They are similar in terms of vocabulary and grammar. AppTek’s highly translation model uses neural network processing to exploit these similarities, enabling sophisticated cross-language translation that is contextually correct and highly accurate. This is especially important for less common, low-resource language pairs where few human-translated texts exist, such as Macedonian-to-English.
The AppTek system also easily handles linguistic "code switching," where two languages, e.g. Ukrainian and Russian, may be mixed within a single sentence, usually a colloquial message, text or spoken.AppTek’s multilingual NMT system can also be fine-tuned to customer domains (e.g. medical texts) in one or more Slavic languages. Further, the system can be integrated with Apptek’s speech recognition solutions so that, for example, Russian speech can be recognized and translated using the same system that translates Ukrainian or Polish.
Dr. Evgeny Matusov, lead science architect for machine translation at AppTek, explained that "Because the focus of most machine translation development is on popular language pairs, other less popular languages often don’t benefit from the latest technology developments such as NMT; consequently those languages still require more expensive and hard-to-find human translation. At AppTek, we solve this problem by creating a single multilingual NMT system that is able to deliver high-quality translation of both high- and low-resource languages."
Posted by Veronica Silva Cusi, news correspondent
Date Posted: Thursday, January 24, 2019
ContactCenterWorld.com is the world's premier on-line resource for the call and contact center industry. This article is one of hundreds available on-line to registered members. Our resource is updated every working day and includes content from every corner of the world. If you are not a registered member go to www.ContactCenterWorld.com and register today.