NATURAL LANGUAGE PROCESSING FOR PUBLIC FEEDBACK ANALYSIS: UNCOVERING CITIZEN SENTIMENTS IN POLICY IMPLEMENTATION IN THE UNITED STATES

Abstract

The research explores the ability of Natural Language Processing (NLP) to undertake systematic analysis of citizen views on policy delivery. Quantitative-computational study design with qualitative underpinning was utilized, providing room for large-scale computational examination as well as contextual understanding of citizens' voices. Data were gathered from social media, government consultation websites, and online forums, using purposive sampling to identify items of particular applicability to active policies. The text data was subjected to aggressive preprocessing in the form of normalization, tokenization, lemmatization, stop-word removal, and translation of non-English comments. Analytical methods integrated sentiment analysis, topic modelling based on LDA, BER Topic, and supervised classification via conventional machine learning models, as well as cutting-edge transformer-based deep learning models like BERT. Outcomes indicate that citizens were overall satisfied in policy areas like the economy and health, but dissatisfactions were heavily felt in service delivery, infrastructure, and openness of policy. Quantitatively speaking, 42.8% of the comments were positive, 28.7% neutral, and 28.5% negative, and notable themes were service delivery (31%) and openness of policy (24%). Comparative analysis also identified effective alignment of citizens' issues and the government's reactions within economic and health domains, but ineffective alignment within transparency and infrastructure domains. Performance assessment indicated that BERT performed best among other models with a 90.6% F1-score, which accounts for its capacity to identify subtle sentiments from citizens. The findings confirm the significance of NLP as a tool for policy analysis, having the ability to detect inconsistency in policy action and citizen expectations. By combining computational observations with qualitative assessment, this research also highlights the capacity of AI-based feedback analysis to enhance the responsiveness, inclusivity, and transparency of the government.