Extracción de Palabras Clave de Ciberacoso de Textos Breves: un Enfoque de Aprendizaje Automático

Cyberbullying has a negative impact on society due to the consequences suffered by victims, bullies, and bystanders. Widespread access to the internet and social networks, especially among young people without the tools to deal with these situations, makes social education necessary to mitigate the effects of cyberbullying. This study seeks to contribute to this training through the creation of scripts for educational capsules. To this end, a model was developed that automates the search and extraction of data from the social network X using Python and Selenium Web Driver. After a text preprocessing process using Natural Language Processing techniques, the Latent Dirichlet Assignment (LDA) model was applied to identify keywords. Finally, the pre-trained model "text-davinci-003" was used through the OpenAI API to generate the content of the educational capsules, assigning a context and using the identified keywords. The outcome of this proposed research is the generation of a script that includes topics on education and the prevention of bullying and cyberbullying. To ensure the reliability of the text generated by the pre-trained generative model, it was evaluated by an expert in the field using the Goal-Question-Metric (GQM) approach, which validates its potential in generating educational content in the fight against cyberbullying.