Company’s Results in High-Profile Evaluation Campaign
Showcase World-Class Spoken Language Translation Technology Solutions
MCLEAN, Va.--(BUSINESS WIRE)--#ASR--AppTek, a leader in Artificial Intelligence (AI), Machine Learning (ML), Automatic Speech Recognition (ASR), Neural Machine Translation (NMT) and Natural Language Understanding (NLU) technologies, today announced that its platform ranked first in multiple categories of the evaluation campaign of the 17th annual International Workshop on Spoken Language Translation (IWSLT 2020). The competition was held in March and April of 2020, with results announced during a July workshop hosted by the Association for Computational Linguistics.
AppTek entered the competition to benchmark the performance of its speech technology platform, developed under the direction of AppTek’s Head of Scientific Research and Development Prof. Dr.-Ing. Hermann Ney, against other industry-leading academic institutions and science teams. The company participated in tracks for English-to-German speech translation, out of which it won first place in Offline Speech Translation and Non-Native Speech Translation. The platform achieved best-in-class rankings in measurements including Word Error Rate (WER) for ASR, and Bilingual Evaluation Understudy (BLEU) and Translation Error Rate (TER) scores for MT.
Using a model close to production settings, AppTek achieves the best possible speech translation quality for German when measured against its academic and enterprise peers. That included completing a task with a non-native English source. In addition, the company’s end-to-end neural model that directly translates English speech into German text reached the quality level of the cascaded approach with separate ASR and MT components, further highlighting AppTek’s speech translation performance.
"We are thrilled that AppTek was first not only in offline speech translation evaluations for both native and non-native English speech, but also received exemplary marks with low latency in the simultaneous translation track. It is a novel research area, and our original approach devised and developed over the past year has proven to produce almost immediate partial translations of spoken utterances on-the-fly,” said Evgeny Matusov, Lead Science Architect, Machine Translation, at AppTek. “AppTek’s leading results in this track reflect the hard work, skill and expertise of our scientists, including colleagues Patrick Wilken and Tamer Alkhouli. We look forward to delivering these innovations to our customers in the coming year, further cementing our status as the leader in customizable speech technology.”
The full report of IWSLT 2020 results can be found here.
AppTek is a global leader in artificial intelligence (AI) and machine learning (ML) technologies for automatic speech recognition (ASR), neural machine translation (NMT), and natural language understanding (NLU). The AppTek platform delivers industry-leading, real-time streaming and batch technology solutions in the cloud or on-premise for organizations across a breadth of worldwide markets such as media and entertainment, call centers, government, enterprise business, and more. Built by scientists and research engineers who are recognized among the best in the world, AppTek’s solutions cover a wide array of languages, dialects, and channels. For more information, please visit http://www.apptek.com.