Toxicity in Spanish News Comments and its Relationship with Constructiveness

Resumen

Online news comments are a critical source of information and opinion, but they often become a breeding ground for toxic discourse and incivility. Detecting toxicity in these comments is essential to understand and mitigate this problem. This paper presents a corpus of Spanish news comments labeled with toxicity (NECOS-TOX) and conducts a series of experiments using several machine learning algorithms, including different language models based on transformers. Our findings show that Spanish language models, such as BETO, are capable of detecting toxicity in Spanish news comments. Additionally, we investigated the relationship between toxicity and constructiveness in these comments and found that there is no clear correlation between the two factors. These results provide insights into the complexities of online discourse and highlight the need for further research to better understand the relationship between toxicity and constructiveness in Spanish news comments.