A collection of 100 news articles in Marathi language along with their corresponding summaries. The summaries are extractive in nature and have been acknowledged by a language expert.
Results obtained by baseline approaches on this dataset after basic preprocessing steps (stop word removal, suffix stemming) were as follows:
Approach | ROUGE-1 F1 | ROUGE-2 F1 | ROUGE-L F1 |
---|---|---|---|
Position based textrank | 0.646 | 0.592 | 0.659 |
Similarity based textrank | 0.648 | 0.591 | 0.661 |
SVD | 0.612 | 0.512 | 0.626 |