Science mapping via dynamic topic modelling: an analysis on 30 years of Social Indicators Research

Considerable interest has been devoted in recent years to the quantitative study of the scientific literature, thanks to the availability of online resources (e.g. Web of Science) and the development of effective techniques for performing automatic analyses. The procedures commonly implemented in bibliometric studies are the performance analysis and the science mapping. The first one aims at evaluating the literature related to a given domain on the basis of bibliographic data. The second one tries to highlight the structural and cognitive patterns of the domain. Mapping techniques frequently refer to textual data analysis. Each domain or theme can be characterised by a set of keywords, assigned by the authors of the publications or by the indexing services. In this paper, an analysis on the last 30 years of Social Indicators Research (SIR) is performed. Founded in 1974, SIR has become one of the leading journals on problems related to the measurement of the different aspects involving the social sphere. Aiming at describing the evolution over time of the SIR themes, here we refer to the dynamic topic modelling (DTM) approach (Blei & Lafferty, 2006). DTM is an extension of the well-known latent Dirichelet allocation. It captures the temporal evolution of topics in a sequentially organised collection of documents. The notion of time is included using the meta-data of the documents. By applying DTM to the 1987-2017 SIR collection we show the evolution of keyword-themes distribution, underlying the themes of the collection over the time and tracking how they have changed.