Researchers in Ecuador are using deep learning techniques to identify the characteristics of bullying behaviour in Spanish language text on social media systems. Details are provided in the International Journal of Data Mining, Modelling and Management.
Paúl Cumba-Armijos, Diego Riofrío-Luzcando*, Verónica Rodríguez-Arboleda and Joe Carrión-Jumbo Digital School, SEK International University, Quito, Pichincha, Ecuador have extracted expressions and phrases that might commonly be used in episodes of cyberbullying from 83400 updates on one particular social network. They have used this body of text to train a convolutional neuronal network. The algorithm that emerges from this training is a tool that can then autonomically identify insults, racism, homophobic attacks, and so on.
It is perhaps well recognised that although there are huge benefits wrought by social media and social networking tools. However, as with any invention, there are always those who might seek to abuse the system for their own malicious ends. Such activity might involve the further marginalisation of vulnerable groups and young people and so it is desirable to find ways to ameliorate the risk to such groups from cyberbullies. The team writes that in Ecuador, 27% of teenagers have reported suffering marginalisation through cyberbullying, 46% have reported harassment, 17% aggressive behaviour online, and 10% have experienced extortion.
Tests on the trained neural network by the team showed that it works with a high precision of more than 98 percent. The next step, which may well improve that precision, would be to draw in data from blogs and additional social media sites and to incorporate additional Spanish phrases to improve the system’s prediction capabilities.
Cumba-Armijos, P., Riofrío-Luzcando, D., Rodríguez-Arboleda, V. and Carrión-Jumbo, J. (2022) ‘Detecting cyberbullying in Spanish texts through deep learning techniques’, Int. J. Data Mining, Modelling and Management, Vol. 14, No. 3, pp.234–247.
No comments:
Post a Comment