Due to the recent popularity and growth of social media platforms such as Facebook and Twitter, cyberbullying is becoming more and more prevalent. The current research on cyberbullying and the NLP techniques being used to classify this kind of online behaviour was initially studied. This paper discusses the experimentation with combined Twitter datasets by Maryland and Cornell universities using different classification approaches like classical machine learning, RNN, CNN, and pretrained transformer-based classifiers. A state of the art (SOTA) solution was achieved by optimising BERTweet on a Onecycle policy with a Decoupled weight decay optimiser (AdamW), improving the previous F1-score by up to 8.4%, resulting in 64.8% macro F1. Particle Swarm Optimisation was later used to optimise the ensemble model. The ensemble developed from the optimised BERTweet model and a collection of models with varying data representations, outperformed the standalone BERTweet model by 0.53% resulting in 65.33% macro F1 for TweetEval dataset and by 0.55% for combined datasets, resulting in 68.1% macro F1.
Cyberbullying has become a highly problematic occurrence due to its potential of anonymity and its ease for others to join in the harassment of victims. The distancing effect that technological devices have, has led to cyberbullies say and do harsher things compared to what is typical in a traditional face-toface bullying situation. Given the great importance of the problem, detection is becoming a key area of cyberbullying research. Therefore, it is highly necessary for a framework to accurately detect new cyberbullying instances automatically. To review the machine learning and deep learning approaches, two datasets were used. The first dataset was provided by the University of Maryland consisting of over 30,000 tweets, whereas the second dataset was based on the article 'Automated Hate Speech Detection and the Problem of Offensive Language' by Davidson et al., containing roughly 25,000 tweets. The paper explores machine learning approaches using word embeddings such as DBOW (Distributed Bag of Words) and DMM (Distributed Memory Mean) and the performance of Word2vec Convolutional Neural Networks (CNNs) to classify online hate.
Social media has revolutionized the way individuals connect and share information globally. However, the rise of these platforms has led to the proliferation of cyber-hate, which is a significant concern that has garnered attention from researchers. To combat this issue, various solutions have been proposed, utilizing Machine learning and Deep learning techniques such as Naive Bayes, Logistic Regression, Convolutional Neural Networks, and Recurrent Neural Networks. These methods rely on a mathematical approach to distinguish one class from another. However, when dealing with sentiment-oriented data, a more ''critical thinking'' perspective is needed for accurate classification, as it provides a more realistic representation of how people interpret online messages. Based on a literature review conducted to explore efficient classification techniques, this study applied two machine learning classifiers, Multinomial Naive Bayes and Logistic Regression, to four online hate datasets. The results of the classifiers were optimized using bio-inspired optimization techniques such as Particle Swarm Optimization and Genetic Algorithms, in conjunction with Fuzzy Logic, to gain a deeper understanding of the text in the datasets.
Due to the recent popularity and growth of social media platforms such as Facebook and Twitter, cyberbullying is becoming more and more prevalent. The current research on cyberbullying and the NLP techniques being used to classify this kind of online behaviour was initially studied. This paper discusses the experimentation with combined Twitter datasets by Maryland and Cornell universities using different classification approaches like classical machine learning, RNN, CNN, and pretrained transformer-based classifiers. A state of the art (SOTA) solution was achieved by optimising BERTweet on a Onecycle policy with a Decoupled weight decay optimiser (AdamW), improving the previous F1-score by up to 8.4%, resulting in 64.8% macro F1. Particle Swarm Optimisation was later used to optimise the ensemble model. The ensemble developed from the optimised BERTweet model and a collection of models with varying data representations, outperformed the standalone BERTweet model by 0.53% resulting in 65.33% macro F1 for TweetEval dataset and by 0.55% for combined datasets, resulting in 68.1% macro F1.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.