ADVANCEMENTS IN GENDER IDENTIFICATION USING TRANSFORMER MODELS IN SOCIAL MEDIA TEXTS

Abstract

Gender identification in text, especially in social media platforms, poses a significant challenge due to the informal, diverse, and often ambiguous nature of online communication. Traditional methods, such as lexicon-based and machine learning approaches, struggle to accurately classify gender, particularly when handling sarcasm, humor, or non-binary gender identities. This paper proposes an advanced approach using transformer models, specifically BERT (Bidirectional Encoder Representations from Transformers), to improve gender classification accuracy in social media texts.

Our study presents a comparative analysis of transformer-based models with traditional methods, demonstrating superior performance in gender classification tasks. By leveraging the contextual understanding of transformer models, we address the limitations of previous approaches, particularly in recognizing diverse gender identities and handling the intricacies of informal text. Our results show a significant improvement in precision, recall, and overall accuracy, along with a reduced bias in gender classification. Additionally, we propose novel strategies for mitigating model bias and ensuring fair representation of non-binary genders.

This work contributes to the growing body of research in natural language processing (NLP) and gender studies by enhancing gender identification models that are both more accurate and inclusive of gender diversity in online communications.