IJAIEM

Obumneme Ukandu

Abstract

Employee sentiment analysis plays a crucial role in Human Resource Management in understanding workforce satisfaction, engagement, and productivity. Conventional sentiment analysis approaches such as rule-based and classical machine learning models, usually fail to perform context sensitivity and sentiment polarity flip, resulting in poor predictions. In order to overcome these shortcomings, the present paper introduces a hybrid deep model that combines Bidirectional Long Short-Term Memory and an Attention Mechanism for improved sentiment classification in HRM.BiLSTM efficiently obtains long-distance dependencies in text reviews, whereas the Attention Mechanism enhances interpretability by highlighting essential words. Experimental test results on an Employee Review dataset show that the new model attains an accuracy of 98.4%, F1-score of 97.6%, and Precision-Recall AUC of 98.2%, surpassing CNN and baseline LSTM models. The findings reveal that BiLSTM- Attention greatly enhanc