AppTek Launches New Metadata-Based Neural Machine Translation System for Enterprises; Extends MT language and dialect coverage

The new state-of-the-art system offers enterprise clients and translation professionals advanced customization options for multi-domain, multi-dialect and multi-genre translations, which improve accuracy and further accelerate translation and localization workflows.

MCLEAN, Va., April 14, 2022 /PRNewswire/ — AppTek, a leader in artificial intelligence (AI) and machine learning (ML) for automatic speech recognition (ASR), neural machine translation (NMT), speech processing/understanding natural language (NLP/U) and synthesis-Speech (TTS) technologies, today announced the release of its new neural machine translation system that embeds metadata as inputs used to personalize MT output and give localization of more accurate machine translations influenced by the user. Additionally, the company has extended its core machine translation platform to support hundreds of language pairs and dialects.

AppTek’s new meta-aware NMT system changes the paradigm of how professional translators work with machine translation output. Until today, most standard machine translation systems operated inside a “black box” where source language text is formulated into target language text without any knowledge or limited by the surrounding context or the domain or topic of the source text. , and with limited control of the resulting output. Traditionally, companies had to train, deploy and maintain multiple machine translation systems to accommodate translation tasks that differ in aspects such as language, dialect, domain, subject, etc., at the risk of costly translation costs. high deployment and overfitting models.

With AppTek’s new metadata-based NMT platform, enterprise customers can now access a single NMT system with cross-domain, cross-genre, and cross-dialect content, increasing quality and adaptability of the system. By introducing additional metadata into the system, they gain more control over the MT output and can allow translators to simply “flip the switch” to the desired custom translation through relevant features in the editing tools UI. with which professionals work.

Here are examples of MT output customization achieved with the use of additional metadata:

  • Style – switch between formal and informal styles, such as between a telenovela and a documentary, and obtain a translation with an appropriate register of politeness according to the status and relations of the speaker;

  • Length control for automatic dubbing and subtitling tasks – generate shorter or longer translations with minimal loss or distortion of information for tasks with strict length constraints;

  • Speaker Gender – switching to the right kind of speaker, which influences inflections for certain parts of speech, especially in morphologically rich languages ​​like Czech;

  • Domain – adapt to the genre of the text, such as news programs, patents, talk shows, etc. to increase the overall accuracy and use of relevant translations in the domain of ambiguous words at the document level;

  • Extended context – optionally have the system consider neighboring sentences in a document when translating a particular sentence so that ambiguity, for example, in the translation of pronouns can be resolved.

  • Glossary – take into account official or mandatory translations that the system might otherwise translate differently; and,

  • Linguistic variety – take into account several languages ​​and dialects within the same system, as well as manage the content in several languages.

“By incorporating metadata to influence MT output, we are able to inject ‘global knowledge’ into our platform,” said Yevgeny Matousov, AppTek’s Principal Science Architect for Neural Machine Translation. “This improves the overall quality and adaptability of system output and can be accomplished in a single, versatile system designed to reduce environmental footprint and cost.”

AppTek’s metadata-driven machine translation technology is now available for translation from English to select European languages ​​and their varieties, with more language pairs coming soon. The system can be customized and adapted to the needs of corporate clients using existing parallel domain-specific translation corpora found in corporate archives.

“As the demand for content localization continues to skyrocket, companies must continue to innovate and find new ways to further accelerate production workflows,” said Kyle Maddock, SVP Marketing at AppTek. “Our metadata-based machine translation system has been specifically designed for translation professionals, giving them greater control over machine translation output, which can further speed up the localization process.”

In addition to its metadata-informed NMT system, AppTek has also extended its core MT platform to cover a long list of languages ​​and dialects, including the addition of Indic and Slavic languages. It now supports Afrikaans, Albanian, Amharic, Arabic (multidialect), Armenian, Azeri, Bengali, Belarusian, Bosnian, Bulgarian, Catalan, Chinese ( multidialect), Croatian, Czech, Danish, Dari, Dutch, English (multidialect). -dialect), Estonian, Farsi, Finnish, French (multidialect), Georgian, German, Greek, Gujarati, Hausa, Hebrew, Hindi, Hungarian, Icelandic, Indonesian, Italian, Japanese, Kannada, Kazakh, Korean, Kyrgyz, Latvian, Lithuanian , Macedonian, Malay, Malayalam, Marathi, Mongolian, Norwegian, Pashto, Polish, Portuguese (multidialect), Punjabi, Romanian, Russian, Serbian, Slovak, Slovenian, Somali, Spanish (multidialect), Swedish, Tagalog, Tamil, Telugu, Tigrinya , Thai, Turkish, Turkmen, Ukrainian, Urdu and Uzbek.

For more information, visit

About AppTek
AppTek is a global leader in artificial intelligence (AI) and machine learning (ML) technologies for automatic speech recognition (ASR), neural machine translation (NMT), natural language processing/understanding (NLP) /U) and voice synthesis (TTS). The AppTek platform provides industry-leading real-time and batch streaming technology solutions in the cloud or on-premises for organizations in a wide range of global markets such as media and entertainment, centers calls, government, business, etc. Designed by world-renowned scientists and research engineers, AppTek’s multi-dimensional 4D solutions for HLT (Human Language Technology) with slice and dice methodology spanning hundreds of languages/dialects, domains, channels and demographics drive high-impact results with speed and accuracy. For more information, please visit‍

Media Contact:
Kyle Maddock
[email protected]


Show original content: html