Biological gender identification in Turkish news text using deep learning models

Abstract

Identifying the biological gender of authors based on the content of their written work is a crucial task in Natural Language Processing (NLP). Accurate biological gender identification finds numerous applications in fields such as linguistics, sociology, and marketing. However, achieving high accuracy in identifying the biological gender of the author is heavily dependent on the quality of the collected data and its proper splitting. Therefore, determining the best-performing model necessitates experimental evaluation. This study aimed to develop and evaluate four learning algorithms for biological gender identification in news texts. To this end, a comprehensive dataset, IAG-TNKU, was created from a Turkish newspaper, comprising 43,292 news articles. Four models utilizing popular machine learning algorithms, including Naive Bayes and Random Forest, and two deep learning algorithms, Long Short Term Memory and Convolutional Neural Networks, were developed and evaluated rigorously. The results indicated that the Long Short Term Memory (LSTM) algorithm outperformed the other three models, exhibiting an exceptional accuracy of 88.51%. This model's outstanding performance underpins the importance of utilizing innovative deep learning algorithms for biological gender identification tasks in NLP. The present study contributes to extant literature by developing a new dataset for biological gender identification in news texts and evaluating four machine learning algorithms. Our findings highlight the significance of utilizing innovative techniques for biological gender identification tasks. The dataset and deep learning algorithm can be applied in many areas such as sociolinguistics, marketing research, and journalism, where the identification of biological gender in written content plays a pivotal role. © The Author(s), under exclusive licence to Springer Science+Business Media, LLC, part of Springer Nature 2023.