Natural language processing for data analysis: an application on the well-being

This article presents a case study on the analysis of semi-structured interviews conducted in Italy, using a small dataset. Two supervised approaches are applied to identify key questions to retrieve information and compare responses to the same questions across different respondents. The first approach is based on a bag-of-words model, while the second relies on embeddings. These approaches are compared with two topic modeling methods (LDA and BERTopic). The results highlight the differences between the methods: key-question-based approaches seem to be more suitable when the goal is to compare responses to specific questions, whereas topic modeling techniques are better suited for identifying latent topics.