Argument Aspect Corpus - Nuclear Energy

The Argument Aspect Corpus–Nuclear Energy (AAC-NE) contains English-language sentences with aspect annotations describing the content of arguments on the topic of nuclear energy. It was introduced in this paper: Jurkschat, L., Wiedemann, G., Heinrich, M., Ruckdeschel, M., &amp; Torge, S. (2022). Few-Shot Learning for Argument Aspects of the Nuclear Energy Debate. In Proceedings of the 13th International Conference on Language Resources and Evaluation (LREC 2022). European Language Resources Association (ELRA). The AAC-NE corpus is based on a subset of all argumentative sentences contained in the UKP SAM dataset [1] for which a majority vote of three annotators could be achieved during the annotation of the main argument aspect of each sentence. The CSV files contain one of nine aspect labels per argumentative sentence split into training, dev, and test set. <strong>aspect</strong> <strong>train</strong> <strong>dev</strong> <strong>test</strong> <strong>Sum</strong> <strong>Kripp. Alpha</strong> alternatives 100 16 21 137 0.69 costs 98 17 29 144 0.72 environment 209 27 64 300 0.74 innovation 33 2 8 43 0.38 reactor safety 112 17 43 172 0.59 reliability 47 5 10 62 0.36 waste 87 5 26 118 0.80 weapons 52 11 15 78 0.77 other 120 23 29 172 0.49 <strong>all</strong> <strong>858</strong> <strong>123</strong> <strong>245</strong> <strong>1226</strong> <strong>0.62</strong> pro 706 cons 520 Additionally, it contains 2000 unlabeled sentences with presumably argumentative content sampled from the newspaper “The Guardian”. [1] Stab, C., Miller, T., Schiller, B., Rai, P., &amp; Gurevych, I. Cross-topic Argument Mining from Heterogeneous Sources. In E. Riloff, D. Chiang, J. Hockenmaier, &amp; J. Tsujii (Eds.), Proceedings of the 2018 Conference on Empirical Methods in Natural Language Processing (pp. 3664–3674). Association for Computational Linguistics. https://doi.org/10.18653/v1/D18-1402