You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
BertScore Implementation for TextGeneration in NLP
Motivation
BERTScore computes a similarity score for each token in the candidate sentence with each token in the reference sentence. However, instead of exact matches, we compute token similarity using contextual embeddings.
Pitch
need an api like bert_score(translate_corpus, reference_corpus, ...)
🚀 Feature
BertScore Implementation for TextGeneration in NLP
Motivation
BERTScore computes a similarity score for each token in the candidate sentence with each token in the reference sentence. However, instead of exact matches, we compute token similarity using contextual embeddings.
Pitch
need an api like
bert_score(translate_corpus, reference_corpus, ...)
Related Paper: https://arxiv.org/abs/1904.09675
A Really Good Explanation: https://wiki.math.uwaterloo.ca/statwiki/index.php?title=BERTScore:_Evaluating_Text_Generation_with_BERT
Alternatives
This is an implementation: https://github.com/Tiiiger/bert_score
The text was updated successfully, but these errors were encountered: