Closed zwhe99 closed 1 year ago
@neubig https://github.com/zwhe99/gpt-MT also has outputs of DeepL. However, DeepL does not support en<->ha and en<->is. I'm not sure how to handle a situation where a system only supports partial language pairs
Thanks a bunch @zwhe99! I'll take a look and play around with this now.
To answer your question, it's already possible to create slices for each language pair to investigate accuracy be language pair separately, and we could create a slice "DeepL Supported Language Pairs" for the average over all language pairs that DeepL supports as well.
Thank you! I hope this helps the community!
Description
This PR adds system outputs of [google-cloud, gpt-3.5-turbo-0301, gpt-4-0314] to MT analysis example.
examples/analysis_gpt_mt/config.py
examples/analysis_gpt_mt/README.md
References
128