Sentiment Analysis of Work-From-Home Practices Among Indian IT Professionals using Transformer-Based Models

Work From Home changed the face of the IT employee in India from being confined to working from the office to working from home which has its pros and cons regarding productivity, work-life balance, and stress levels. This research uses Transformer-based deep learning models such as BERT and RoBERTa for sentiment analysis concerning employees’ ideas on WFH. When applied to a new data science set specifically designed for Indian B2B IT professionals, the proposed model had 90.2% accuracy, 89.1% precision, and $89.1 \%$ F1 score and was better than traditional models. Thus, it provides a strong foundation for the representation of strongly context-dependent sentiments in many languages, rectifying weaknesses in prior research. The outcomes presented a positive attitude towards productivity and work and work-life balance but revealed apprehension with mental health. It can help organizations in designing new progressive and flexible work-from-home policies to enhance the work ambiance. Future work concerns the extension of the set of used datasets and the investigation of continuous sentiment analysis for more fields of use.