SOTA machine translation at Unbabel
Practical AI
00:00
Advancing Translation Evaluation: Beyond BLEU to Comet
This chapter examines the evolution of translation evaluation metrics, contrasting traditional methods like BLEU with newer approaches like Comet. It highlights how Comet utilizes advanced language model representations and human feedback to enhance translation assessment accuracy. Additionally, the chapter discusses the challenges in machine translation, including quality estimation methodologies and the practical difficulties of deploying large models.
Transcript
Play full episode
