Meta Claims It Just Made ‘Speech Translation a Whole Lot Better’
Built on the August 2023 SeamlessM4T, the Seamless suite includes models that preserve vocal style and translate text and speech in nearly 100 languages.
*New* Slator Pro Guide: Scaling an LSP Key Account — Growing Small Clients into Key Accounts
News and analysis of the latest developments in machine translation, computer-aided-translation, natural language processing, and other language-related areas in artificial intelligence.
Built on the August 2023 SeamlessM4T, the Seamless suite includes models that preserve vocal style and translate text and speech in nearly 100 languages.
New research reveals the best-performing machine translation evaluation metrics, identifies major challenges in metrics development, and suggests improvements.
New research suggests that integrating quality metrics as reward models into the machine translation pipeline can enhance the quality of the generated text.
New research shows that large language models are better translators of Arabic dialects than commercial machine translation systems but remain far from perfect.
Toronto-based client communication platform Messagepoint announced it will integrate both DeepL and OpenAI machine translation capabilities in its content hub starting November 30, 2023.
Researchers from Shanghai Jiao Tong University and Tencent AI Lab introduce a method to elevate word-level auto-completion through machine translation, with experimental results showcasing noteworthy enhancements.
A study introduces an approach to streamline translation between related languages, with the goal of enhancing trade efficiency and strengthening social connections, particularly in regions with related languages.
Microsoft Azure AI researchers explore the potential of large language models for automatic post-editing and find that LLMs are good but not great at it.
Amid a deluge of research into speech translation, many of the world’s leading technology companies are teaming up with academia to accelerate progress in this challenging field of language AI.
Brown University researchers reveal an issue with AI safety mechanisms in large language models involving low-resource languages.
A study demonstrates the ability of large language models to remove noise from datasets and underscores their potential for data cleaning.
Carnegie Mellon University researchers explore LLM effectiveness across 204 languages revealing their output limitations for low-resource languages.
Interest in automating machine translation quality estimation increases with the prevalence of large language models, but still has some way to go to be deployed at scale.
As large language models provide ever more sophisticated ways to automatically generate content, the potential for prompting to improve machine translation output becomes clearer.
The World Intellectual Property Organization unveiled its in-house solution designed to generate conference meeting transcripts and machine translations.
Monash University researchers show that large language models can do real-time machine translation and propose new ways for model fine-tuning.
Google created a new dataset for machine translation and multilingual NLP tasks across 400 languages and released a high-performing multilingual MT model trained on this data.
Researchers from NVIDIA, Factored.ai, Talon Voice, and others open-source a properly licensed dataset of 1,780 hours of speech in 77 different languages, plus transcriptions.
Meta introduced SeamlessM4T after months of research and numerous expansive multilingual models. How does the “next big thing” fit into future projects, and the language industry?
G/O Media fires Gizmodo’s entire staff of Spanish writers and begins publishing raw machine-translated versions of English articles that are now being indexed by Google Search.
Slator Weekly: Join over 15,800 subscribers and get the latest language industry intelligence every Friday
Tool Box: Join 10,000 subscribers for your monthly linguist technology update.
Your information will not be shared with third parties. No Spam.
This will close in 0 seconds