Gender Classification from Tweets Using LSTMs and Transformers
Omar Alaaeldein · 2022 2nd International Mobile, Intelligent, and Ubiquitous Computing Conference (MIUCC) · 2022
With the rise of social media platforms and personalised Web 2.0, the study of sociolinguistic factors in text such as gender bias has had a huge influx of new data from users that has since been used by consumer-based organizations for data analytics to streamline their digital models and cluster their customers to better cohort their services, and by social scientists and linguistics researchers to further improve their hypotheses and work. In this work, I have gathered data from the social media platform “Twitter”, and used existing data from the web and preprocessing it to train and develop 7 different models for Gender Classification based on short snippet of text using LSTMs and Transformers in conjunction with other techniques including transfer learning using feature extraction and fine-tuning. The results and statistics of the models are discussed, and illustrated in confusion matrices, and a discussion/comparison between the 7 models is performed.