%0 Conference Paper %1 bommasani-cardie-2020-intrinsic %A Bommasani, Rishi %A Cardie, Claire %B Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP) %C Online %D 2020 %I Association for Computational Linguistics %K dataset intrinsic quality %P 8075--8096 %R 10.18653/v1/2020.emnlp-main.649 %T Intrinsic Evaluation of Summarization Datasets %U https://www.aclweb.org/anthology/2020.emnlp-main.649 %X High quality data forms the bedrock for building meaningful statistical models in NLP. Consequently, data quality must be evaluated either during dataset construction or *post hoc*. Almost all popular summarization datasets are drawn from natural sources and do not come with inherent quality assurance guarantees. In spite of this, data quality has gone largely unquestioned for many of these recent datasets. We perform the first large-scale evaluation of summarization datasets by introducing 5 intrinsic metrics and applying them to 10 popular datasets. We find that data usage in recent summarization research is sometimes inconsistent with the underlying properties of the data. Further, we discover that our metrics can serve the additional purpose of being inexpensive heuristics for detecting generically low quality examples.

@inproceedings{bommasani-cardie-2020-intrinsic, abstract = {High quality data forms the bedrock for building meaningful statistical models in NLP. Consequently, data quality must be evaluated either during dataset construction or *post hoc*. Almost all popular summarization datasets are drawn from natural sources and do not come with inherent quality assurance guarantees. In spite of this, data quality has gone largely unquestioned for many of these recent datasets. We perform the first large-scale evaluation of summarization datasets by introducing 5 intrinsic metrics and applying them to 10 popular datasets. We find that data usage in recent summarization research is sometimes inconsistent with the underlying properties of the data. Further, we discover that our metrics can serve the additional purpose of being inexpensive heuristics for detecting generically low quality examples.}, added-at = {2021-01-20T09:54:11.000+0100}, address = {Online}, author = {Bommasani, Rishi and Cardie, Claire}, biburl = {https://www.bibsonomy.org/bibtex/23d9ca3c984b5a9bf7d36caa8593d6bf2/parismic}, booktitle = {Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing (EMNLP)}, description = {Intrinsic Evaluation of Summarization Datasets - ACL Anthology}, doi = {10.18653/v1/2020.emnlp-main.649}, interhash = {d5eb6163f463f5d0370ecebc568fb08a}, intrahash = {3d9ca3c984b5a9bf7d36caa8593d6bf2}, keywords = {dataset intrinsic quality}, month = nov, pages = {8075--8096}, publisher = {Association for Computational Linguistics}, timestamp = {2021-01-20T09:54:11.000+0100}, title = {Intrinsic Evaluation of Summarization Datasets}, url = {https://www.aclweb.org/anthology/2020.emnlp-main.649}, year = 2020 }