Multi-Class Detection: Learning Algorithms on Categories of Cyberbullying | Intro to ML for B&I
The pervasive nature of the internet is an enabler of Cyberbullying, which is commonplace across various social media platforms. Bullying over the internet is a deliberate act of inflicting harm on targeted victims, and threats (harassment) usually come in the form of disrespectful or threatening, or sensitive messages that comes in as either a video, image, voice or textual data from the aggressor. Predicting cyberbullying earlier enough before it can afflict harm on its victims is a wicket problem. Given the complex dimensions in which cyberbullying may occur and the distributed nature of social media. A more sophisticated language mechanism is needed to correctly classify these textual/images/video data in various categories to avoid false alarms. Recent advances in natural language processing (NLP), such as the large language models, are an icebreaker to detect cyberbullying across different languages, cultures, religions, ages, genders, etc. Also, researchers across machine learning communities have been working. They have developed complex deep learning architectures such as transformers (BERT) that can predict complicated NLP tasks, which traditional machine learning models would be limited. A typical BERT model, can contextually generate text embedding for a multi-class problem and task-specific classification embedding. This project implements machine and deep learning models to analyze and compare multi-class prediction problems on cyberbullying. By using a multinominal naive Bayes vs. a logistics regression classifier, a deep neural network (NN), recurrent (RNN), multi-layer perceptron, a long and short-term memory (LSTM) model, and finally, a BERT model on an imbalance dataset of over 47K tweets. | Armstrong Foundjem, PhD program for Computing at Queen's University | Intro to ML for Black & Indigenous Students - Capstone Winner
Смотрите видео Multi-Class Detection: Learning Algorithms on Categories of Cyberbullying | Intro to ML for B&I онлайн без регистрации, длительностью часов минут секунд в хорошем качестве. Это видео добавил пользователь Vector Institute 17 Февраль 2023, не забудьте поделиться им ссылкой с друзьями и знакомыми, на нашем сайте его посмотрели 206 раз и оно понравилось 3 людям.