%0 Conference Paper %1 rony-etal-2022-rome %A Rony, Md Rashad Al Hasan %A Kovriguina, Liubov %A Chaudhuri, Debanjan %A Usbeck, Ricardo %A Lehmann, Jens %B Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) %C Dublin, Ireland %D 2022 %E Muresan, Smaranda %E Nakov, Preslav %E Villavicencio, Aline %I Association for Computational Linguistics %K Evaluating Generation Language Metric Natural Robust %P 5645--5657 %R 10.18653/v1/2022.acl-long.387 %T RoMe: A Robust Metric for Evaluating Natural Language Generation %U https://aclanthology.org/2022.acl-long.387/ %X Evaluating Natural Language Generation (NLG) systems is a challenging task. Firstly, the metric should ensure that the generated hypothesis reflects the reference`s semantics. Secondly, it should consider the grammatical quality of the generated sentence. Thirdly, it should be robust enough to handle various surface forms of the generated sentence. Thus, an effective evaluation metric has to be multifaceted. In this paper, we propose an automatic evaluation metric incorporating several core aspects of natural language understanding (language competence, syntactic and semantic variation). Our proposed metric, RoMe, is trained on language features such as semantic similarity combined with tree edit distance and grammatical acceptability, using a self-supervised neural network to assess the overall quality of the generated sentence. Moreover, we perform an extensive robustness analysis of the state-of-the-art methods and RoMe. Empirical results suggest that RoMe has a stronger correlation to human judgment over state-of-the-art metrics in evaluating system-generated sentences across several NLG tasks.

@inproceedings{rony-etal-2022-rome, abstract = {Evaluating Natural Language Generation (NLG) systems is a challenging task. Firstly, the metric should ensure that the generated hypothesis reflects the reference`s semantics. Secondly, it should consider the grammatical quality of the generated sentence. Thirdly, it should be robust enough to handle various surface forms of the generated sentence. Thus, an effective evaluation metric has to be multifaceted. In this paper, we propose an automatic evaluation metric incorporating several core aspects of natural language understanding (language competence, syntactic and semantic variation). Our proposed metric, RoMe, is trained on language features such as semantic similarity combined with tree edit distance and grammatical acceptability, using a self-supervised neural network to assess the overall quality of the generated sentence. Moreover, we perform an extensive robustness analysis of the state-of-the-art methods and RoMe. Empirical results suggest that RoMe has a stronger correlation to human judgment over state-of-the-art metrics in evaluating system-generated sentences across several NLG tasks.}, added-at = {2025-01-07T13:07:03.000+0100}, address = {Dublin, Ireland}, author = {Rony, Md Rashad Al Hasan and Kovriguina, Liubov and Chaudhuri, Debanjan and Usbeck, Ricardo and Lehmann, Jens}, biburl = {https://puma.scadsai.uni-leipzig.de/bibtex/2a21d7f78bff70d75889ec0e7fe6a94e8/scadsfct}, booktitle = {Proceedings of the 60th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)}, doi = {10.18653/v1/2022.acl-long.387}, editor = {Muresan, Smaranda and Nakov, Preslav and Villavicencio, Aline}, interhash = {78431952038987841c369ad001fc4915}, intrahash = {a21d7f78bff70d75889ec0e7fe6a94e8}, keywords = {Evaluating Generation Language Metric Natural Robust}, month = may, pages = {5645--5657}, publisher = {Association for Computational Linguistics}, timestamp = {2025-01-07T13:07:03.000+0100}, title = {{R}o{M}e: A Robust Metric for Evaluating Natural Language Generation}, url = {https://aclanthology.org/2022.acl-long.387/}, year = 2022 }