Mapping the field of word association research using text mining approach

Word associations (WA) have long been the object of researcher's attention. Initially, they were used in psychology, but now they are widely applied in a wide range of disciplines - from text mining to automatic creativity assessment. However, to the best of our knowledge, to date no attempts have been made to map the interdisciplinary research field related to word association (both as the object and methodology). A large number of research papers on the subject makes a detailed manual literature analysis unrealistic, but vital for using text mining methods. This paper is the first one to have applied structural topic modelling to map the interdisciplinary word association research field. We exported abstracts of the papers related to word associations and published in the period from 2003 to 2022 from the Scopus database and designed topic models with the year of publication and country of the corresponding authors as covariates. This allowed us not only to reveal the major academic topics/latent themes in the word association research area but also to analyze the dynamics of scientific interest to particular topics as well as to establish preferences in topics related to the countries where researchers work. Our results indicate the existence of a wide variety of important research foci in the domain of word association. We revealed 30 topics which were divided into four clusters reflecting the interdisciplinary nature of this object/methodology: 1) WA as a diagnostic tool for cognitive/emotional impairment; 2) WA as a methodology to study cognitive processes related to language production; 3) WA processing and applications related to computer science and NLP; 4) WA as a tool for studying the conceptual structure of an individual. Text mining approach for WA as well as most of the topics from cluster 4 were shown to witness an upward trend. The analysis allowed us to revealed two groups of countries with respect to the type of topic distribution: one with a clear preference for several topics and the other one with a diverse range of topics. Taken together, our findings related to WA research mapping could help scientists - both novice and seasoned ones with different backgrounds – to get a better understanding of the possible applications of this powerful methodology and directions of the study of the phenomenon at hand.