Abstract

MUNETON-SANTA, Guberney; PEREZ-AGUIRRE, Carlos Andrés  and  OROZCO-ARROYAVE, Juan Rafael. Mapping Poverty and Well-Being Through Natural Language Processing. rev.estud.soc. [online]. 2025, n.93, pp.39-65.  Epub Aug 25, 2025. ISSN 0123-885X.  https://doi.org/10.7440/res93.2025.03.

Poverty and well-being indexes encompass dimensions that capture meaningful aspects of life worth measuring. These dimensions reflect people’s concerns and priorities, offering insight into their lived experiences. This study identifies dimensions of poverty and well-being directly from people’s everyday language and proposes a novel method for assigning weights to these dimensions based on what people express as important. Using topic modeling techniques within the Natural Language Processing framework, we uncovered key themes people associate with poverty and well-being in their own words. We also applied transfer learning through a zero-shot classification model to assign weights to these dimensions, ranking them by their relevance to the target population studied. In our case studies, the most prominent poverty-related dimensions included lack of opportunities, unemployment, lack of spirit, lack of money, and attempts to get ahead, while the top well-being dimensions identified were living well, meeting basic needs, nourishment, and health. This approach helps pinpoint priority areas for intervention and resource allocation. We recommend using topic modeling techniques when designing multidimensional indicators, as this enables researchers and policymakers to ground social indicators in the voices of the people they aim to serve.

Keywords : artificial intelligence; capability approach; multidimensional poverty; Natural Language Processing; topic modeling; well-being.

        · abstract in Spanish | Portuguese     · text in English     · English ( pdf )