IMPROVING THE AUTOMATIC IDENTIFICATION OF AGGRESSIVE COMMENTS ON SOCIAL MEDIA BY INCORPORATING AUTHOR AND MESSAGE CONTEXT

CASAVANTES MORENO, MARCO EMANUEL (2020) IMPROVING THE AUTOMATIC IDENTIFICATION OF AGGRESSIVE COMMENTS ON SOCIAL MEDIA BY INCORPORATING AUTHOR AND MESSAGE CONTEXT. Maestría thesis, UNIVERSIDAD AUTÓNOMA DE CHIHUAHUA.

Vista Previa

Text Tesis.pdf Download (3334Kb) | Vista Previa

Resumen

In 2019, research was conducted with the purpose of mitigating the problems that involve cyberbullying and hate speech, this work was carried out in Chihuahua and Puebla. People can deal with negative experiences in social networks, such as being a target of cyberbullying or exposing themselves to hateful and vulgar content. It is crucial to address the importance of early identification of users who promote hate speech, as this could allow important outreach programs to prevent consequences in real life, such as self-harm or suicide; however, traditional approaches to detect these behaviors only use the short text messages provided in datasets by task organizers and these texts suffer from lack of context. The objective of this research is to design a classification method based on author and text metadata to identify hostile comments on social networks. In order to build new classification methods for this task, it was necessary to extend the available collections of Twitter tagged data, where in addition to having the text of the tweet and its class, it was possible to retrieve context information in the form of tweet and author features. As a result, we found that there are statistically significant differences between the classification reports of the methods that use metadata with respect to the conventional approaches that lack additional information to the text in the majority of datasets tested.

Actions (login required)

Ver Objeto