ARG-tech will be present this year in Singapore at EMNLP 2023 and the 10th Workshop on Argument Mining. Two papers co-authored by ARG-tech members have been accepted to the EMNLP Main Conference and one paper to the Workshop on Argument Mining.
In the EMNLP long paper titled ‘Automatic Debate Evaluation with Argumentation Semantics and Natural Language Argument Graph Networks’ and co-authored by Ramon Ruiz-Dolz, we propose a hybrid method that combines concepts from Computational Argumentation theory and Natural Language Processing to predict the outcome of complete debates in academic debate tournaments.
In the EMNLP short paper titled ‘VivesDebate-Speech: A Corpus of Spoken Argumentation to Leverage Audio Features for Argument Mining’, we publicly release the largest speech corpus containing argument annotations. In addition to the corpus, initial results on the task of argument mining from spoken data are reported, pointing out that acoustic features can be a relevant addition to textual features in the segmentation of argumentative components.
Finally, in the Argument Mining Workshop paper titled ‘Detecting Argumentative Fallacies in the Wild: Problems and Limitations of Large Language Models’ and authored by Ramon Ruiz-Dolz and John Lawrence, we explore the limitations of the existing approaches to identify fallacies with state-of-the-art LLMs. We conduct an error analysis considering specific instances of arguments and rise a discussion in which we suggest the use of more complete models of argumentation that can make possible to overcome the limitations of the state-of-the-art LLMs.
