Analysing Hate Speech against Migrants and Women through Tweets Using Ensembled Deep Learning Model

Area of Science:

• Natural Language Processing
• Computational Social Science
• Artificial Intelligence

Background:

• Rising popularity of social media platforms like Twitter.
• Increased prevalence of demeaning and hateful speech online.
• Growing need for effective sentiment analysis tools to detect online hate speech.

Purpose of the Study:

• To propose a deep learning model for classifying sentiment in tweets.
• To analyze hate speech targeting migrants and women.
• To differentiate between individual and group-based hate speech perpetrators.

Main Methods:

• Utilizing a deep learning model for sentiment classification.
• Implementing word embedding techniques: Global Vector (GloVe), Term Frequency-Inverse Document Frequency (TF-IDF), and transformer-based embeddings.
• Employing a combination of deep learning models: Bidirectional Long Short-Term Memory (BiLSTM), Convolutional Neural Network (CNN), and Multilayer Perceptron (MLP).

Main Results:

• The deep learning model successfully classifies sentiment in tweets.
• The model can distinguish hate speech directed at specific demographics (migrants, women).
• The model is capable of identifying whether hate speech originates from a single user or a group.

Conclusions:

• Deep learning models offer a robust approach to sentiment analysis on social media.
• The proposed model demonstrates effectiveness in identifying and categorizing hate speech.
• This research contributes to developing tools for combating online harassment and promoting safer digital environments.