Assessing Reference-Free Peer Evaluation for Machine Translation
This work addresses the need for more scalable evaluation in machine translation, though it is incremental as it builds on existing reference-free approaches.
The paper tackled the problem of scalable machine translation evaluation by exploring reference-free methods, showing that scaling up a multilingual model can match BLEU performance and is robust across domains and system qualities.
Reference-free evaluation has the potential to make machine translation evaluation substantially more scalable, allowing us to pivot easily to new languages or domains. It has been recently shown that the probabilities given by a large, multilingual model can achieve state of the art results when used as a reference-free metric. We experiment with various modifications to this model and demonstrate that by scaling it up we can match the performance of BLEU. We analyze various potential weaknesses of the approach and find that it is surprisingly robust and likely to offer reasonable performance across a broad spectrum of domains and different system qualities.