The International Workshop on Spoken Language Translation (IWSLT) has long been a prestigious venue for showcasing the latest advancements in spoken language technology. As we approach the 2024 edition of the conference, AppTek is proud to highlight its contributions and achievements, further solidifying its role as a leader in the field. From pioneering subtitling technology to influential keynote addresses, AppTek's involvement in IWSLT 2024 is a testament to its commitment to innovation and excellence.
IWSLT 2024 is collocated with the 62nd Annual Meeting of the Association for Computational Linguistics (ACL 2024) in Bangkok this August. This annual conference is a hub for experts and researchers coming together to discuss cutting-edge research and technologies in spoken language translation and is synonymous with shared tasks where scientific papers and system descriptions are presented.
Among these experts is Prof. Dr.-Ing. Hermann Ney, AppTek’s Director of Science, who serves on the conference's esteemed steering committee. With insights drawn from years of pioneering work in machine translation and speech processing, Prof. Ney is an influential figure of the conference’s agenda, guiding the focus on emerging trends and challenges in spoken language translation.
There are seven shared tasks hosted annually at the conference, subtitling being one of them. The challenge in this track is to balance linguistic accuracy and fluency with the strict formatting constraints of subtitling. AppTek’s Lead Science Architect, Dr. Evgeny Matusov, a former student of Prof. Hermann Ney, played a pivotal role in the original setup of this track two years ago and has served on its organising committee ever since.
The deep involvement of both these leading scientists in the event underscore AppTek’s integral role in shaping the direction of language technology research and its commitment to driving innovation in the field.
AppTek’s participation in the IWSLT 2024 subtitling track consisted of rigorously evaluating the company’s systems on their ability to translate English speech into German and Spanish subtitles across various content domains which included television series (ITV), sports (Peloton) and educational talks (TED).
AppTek participated with up to 3 system variants under unconstrained conditions and consistently ranked among the top performers in all categories and won the first place overall according to the main automatic metrics SubER (Subtitle Error Rate). The company’s system achieved an overall SubER score of 62.02 in the English->Spanish language pair, and 70.34 in English->German across all domains. It was adapted to entertainment content, so the largest improvements were seen in the television series domain, while the results in the sports and educational talks domains were also very competitive.
AppTek’s neural machine translation system was able to fulfil all standard subtitling constrains, showcasing an impressive compliance rate, reaching or nearing 100% for both language pairs in all metrics: characters per line (CPL), lines per block (LPB), and characters per second (CPS). The detailed results shown in the images below demonstrate the system’s adaptability across multiple subtitling domains and its capability to maintain readability and synchronization with the on-screen dialogue, which is essential for an optimal viewer experience.
AppTek’s superior performance can be attributed to several cutting-edge features integrated into its neural machine translation system which, coupled with its adaptability to different content types, make AppTek’s subtitling technology one of the most advanced in the industry.
AppTek’s contributions to IWSLT 2024 extends beyond the company’s participation in the subtitling track. David Thulke, one of the company’s scientists, is set to deliver a keynote speech at a climate-related workshop organized by the Association for Computational Linguistics (ACL). His presentation will explore how natural language processing (NLP) can contribute to addressing climate change, a topic of growing importance as the world grapples with environmental challenges. His keynote reflects AppTek’s commitment to applying its technological expertise to pressing global issues and demonstrate the braoder impact of its innovations beyond traditional language translation and subtitling.
AppTek’s achievements at this year’s IWSLT reaffirm the company’s position at the forefront of language technology evolution. By pioneering new subtitling techniques, driving forward-thinking research, or addressing global issues like climate change, AppTek is not just meeting the current demands of the industry but paving the way for future advancements. With a strong foundation built on cutting-edge technology and a team of world-class experts, the company is poised to continue leading the way in transforming how we communicate across languages and cultures. As we celebrate these accomplishments, we look forward to the future of AppTek’s contributions to IWSLT and the broader field of language technology.
AppTek.ai 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), large language models (LLMs) and text-to-speech (TTS) technologies. The AppTek platform delivers industry-leading solutions for organizations across a breadth of global 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, channels, domains and demographics.
