Abstract
This study presents an attempt to systematically integrate domain knowledge into the algorithmic classification of large text corpora. Taking the societal problem of 'obesity' as case study, we build an ontology for obesity and integrate this ontology into topic modelling algorithms. The idea is that for the purpose of mapping a given societal issue with a large collection of scientific texts, the combination of a domain ontology and topic modelling may facilitate the tasks of delineating the relevant corpus, classifying it into topics and interpreting the resulting topics. Preliminary results show that the topics obtained adding an ontology to topic modelling are more meaningful. Considering the need of aligning the research with societal needs, our work goes into the direction of providing a tool to map the topics research outputs in mapping categories that are easier to interpret from a user's perspective.