NLP and Topic Modeling with LDA, LSA, and NMF for Monitoring Psychosocial Well-being in Monthly Surveys

Abstract

This article presents an approach for assessing psychosocial concerns using Natural Language Processing (NLP) and topic modeling on text data collected in monthly surveys. We processed a dataset containing more than 10,000 entries from two regional public health department, focusing on psychosocial concerns expressed by the population. Using NLP techniques and topic models like Latent Dirichlet Allocation (LDA), Latent Semantic Analysis (LSA), and Non-Negative Matrix Factorization (NMF), we identifed and analyzed recurring themes. The study found that LDA with unigrams performed best, yielding a coherence score of 0.59, while NMF was less effective. Key emerging themes included emotional well-being, stress, and social isolation, which evolved over time, especially during the COVID-19 pandemic. The results demonstrate that these methods can identify emerging issues and provide valuable information for decision making.

Cite this work

No public copy is stored for this record, so only the metadata is cited.

Stored only when both papers are in the corpus — 11 citations exist upstream.

Papers that cite the same references — the relational fallback while embeddings are still backfilling.