Abstract
Gender identification for social user network from user-generated text has significant applications in social computing, advertising, and online safety. Though, traditional classifiers often fail to capture linguistic subtleties and raise ethical concerns about bias and privacy. On this study, we propose a transformer-based framework for gender identification on social network text which highlighting transparency and fairness. We assess multiple transformer architectures, including BERT and RoBERTa, comparing their performance against traditional machine-learning baselines using the facebook/md-gender-bias dataset. The proposed method achieves improved for F1 scores while reducing gender bias through bias-mitigation and explainability techniques. Experimental findings demonstrate the practicality of large language models for gender prediction when combined with ethical constraints and responsible evaluation.