Mixture of Topic Modeling and Network Analysis. The case-study of climate change on Twitter

The paper proposes a semi-automatic labeling of topics extracted with a Topic Model using the tools of Social Network Analysis. The aim is to attach a label to every topic studying the terms-topics network structure. This method performs a semi-automatic topics labelling by using Latent Dirichlet Allocation model, integrating the network approach with topic generative model. LDA allows to extract latent topics and Social Network Analysis' tools permit to delineate the neighborhood of each topic, fostering a stronger interpretation of the meanings of the topics through the analysis of the extracted topics and documents' terms. To better show the joint use of Topic Model and Network Analysis, we present a case-study of how young people feel the climate change through the messages of user @Fridays4future extracted by International Fridays For Future Twitter account.