A Bert2Bert model on VoA Persian Corpus (a medium-sized corpus of 7.9 million words, 2003-2008) generates headlines. The model achieved a 25.30 ROUGE-2 score.

For more detail, please follow the News Headline Generation repo.

Eval results

The following table summarizes the ROUGE scores obtained by the Bert2Bert model.

% Precision Recall FMeasure
ROUGE-1 43.78 45.52 43.54
ROUGE-2 24.50 25.30* 24.24
ROUGE-L 41.20 42.22 40.76

Questions?

Post a Github issue on the News Headline Generation repo.

Downloads last month
0
Hosted inference API
Summarization