The Argument Reasoning Comprehension Task

The stance detection dataset contains about 4.7K comments annotated on the comment-level as taking stance towards a given topic. Also suitable for argument/non-argument classification or experimenting with sarcasm in argumentation.Argument component identification data contains 1,927 stance-taking arguments with annotated premises (reasons) as spans in either a BIO CoNLL format or DKPro-Argumentation UIMA-based format along with their meta-data (debate topic, stance, users, etc.). There are 4,294 premises (reasons) annotated in total.Abstractive argument summarization dataset contains 1,927 arguments and their abstractive summaries. The summaries were created by compiling manually rephrased gists of the annotated reasons and thus are expected to convey the most important information from the argument. All relevant metadata are available as well.Use the following citation if you use any of the code or the data set:@article{Habernal.et.al.2017.argument.reasoning, author = {Habernal, Ivan and Wachsmuth, Henning and Gurevych, Iryna and Stein, Benno}, title = {The Argument Reasoning Comprehension Task}, journal = {arXiv preprint}, year = {2017}, url = {https://arxiv.org/abs/1708.01425}}