| Resumen |
Argument mining is a critical area within artificial intelligence with significant implications for the future of machine learning models. It is widely believed that advances in argument mining will enhance the ability of models to construct more effective arguments in diverse contexts, including educational and political settings. However, existing research predominantly focuses on identifying argument structures without sufficiently considering the nuanced quality dimensions inherent within them. This study addresses this gap by conducting several experiments. Firstly, it evaluates the performance of traditional machine learning models in detecting arguments. Subsequently, the research investigates how selected quality dimensions impact the performances of argument prediction. The methodology leverages BM25 features with a Random Forest model, achieving notable results with an F1-score of 0.88 and a Spearman’s correlation coefficient of 0.73. These outcomes surpass those of previous models such as IBM’s 2019 Arg-ranker and base-Arg-ranker, which utilized Bert embeddings and achieved Spearman’s scores of 0.41 and 0.42 respectively. |
