GENDER BIAS IN LARGE LANGUAGE MODELS AND WORD EMBEDDINGS

Main Article Content

Komilova Nilufar Abdilkadimovna

Abstract

Abstract. This article examines the presence and mechanisms of gender bias in natural language processing (NLP) systems, particularly large language models (LLMs) and word embeddings. It explores how AI systems infer gender from textual data, the ethical implications of such inference, and the ways in which stereotypes are encoded in training datasets. The study synthesizes findings from computational linguistics, machine learning fairness research, and sociolinguistics to explain how bias emerges in data-driven language technologies. It also discusses current approaches for detecting gender stereotypes in datasets and evaluates mitigation strategies aimed at improving fairness in NLP systems.

Article Details

Section

Articles

References

1.Bolukbasi, T. et al. (2016). Man is to Computer Programmer as Woman is to Homemaker? Debiasing Word Embeddings.

2.Caliskan, A., Bryson, J. J., & Narayanan, A. (2017). Semantics derived automatically from language corpora contain human-like biases.

3.Devlin, J. et al. (2019). BERT: Pre-training of Deep Bidirectional Transformers for Language Understanding.

4.Garg, N. et al. (2018). Word embeddings quantify 100 years of gender and ethnic stereotypes.

5.Zhao, J. et al. (2018). Gender bias in coreference resolution systems.

6.Sheng, E. et al. (2019). The woman worked as a babysitter: On biases in language generation.