Display translation benchmark results from NTREX dataset
Manage and label datasets for your projects
Create datasets with FAQs and SFT prompts
Upload files to a Hugging Face repository
Convert PDFs to a dataset and upload to Hugging Face
Access NLPre-PL dataset and pre-trained models
Create a domain-specific dataset project
Data annotation for Sparky
Convert a model to Safetensors and open a PR
Rename models in dataset leaderboard
Search for Hugging Face Hub models
Create a report in BoAmps format
Create and manage AI datasets for training models
TREX Benchmark En Ru Zh is a translation benchmark dataset designed to evaluate machine translation systems between English, Russian, and Chinese. It is part of the NTREX dataset family, focusing on providing high-quality test sets for translation tasks. This benchmark is widely used to assess the performance of machine translation models and improve their accuracy and fluency in these language pairs.
• Multilingual Support: Covers English-Russian (En-Ru), English-Chinese (En-Zh), and Russian-Chinese (Ru-Zh) translation tasks.
• Comprehensive Test Sets: Includes diverse and representative test sentences from various domains.
• Regular Updates: The dataset is updated periodically to reflect real-world language usage and evolving translation challenges.
• Detailed Metrics: Provides evaluation metrics such as BLEU, ROUGE, and METEOR scores to assess translation quality.
• Open Access: Available for research and commercial use, promoting collaboration and innovation in machine translation.
What language pairs are supported by TREX Benchmark En Ru Zh?
TREX Benchmark En Ru Zh supports English-Russian (En-Ru), English-Chinese (En-Zh), and Russian-Chinese (Ru-Zh) translation tasks.
How do I interpret the evaluation metrics?
Metrics like BLEU (higher is better) measure the similarity between your model's output and the reference translation. Lower scores indicate room for improvement.
Where can I find more information about TREX Benchmark En Ru Zh?
Additional details, updates, and documentation can be found on the official NTREX dataset website or academic publications related to the TREX benchmark.