Wide-Scale Automatic Analysis of 20 Years of ITS Research

Abstract

The analysis of literature within a research domain can provide significant value during preliminary research. While literature reviews may provide an in-depth understanding of current studies within an area, they are limited by the number of studies which they take into account. Importantly, whilst publications in hot areas abound, it is not feasible for an individual or team to analyse a large volume of publications within a reasonable amount of time. Additionally, major publications which have gained a large number of citations are more likely to be included in a review, with recent or fringe publications receiving less inclusion. We provide thus an automatic methodology for the large-scale analysis of literature within the Intelligent Tutoring Systems (ITS) domain, with the aim of identifying trends and areas of research from a corpus of publications which is significantly larger than is typically presented in conventional literature reviews. We illustrate this by a novel analysis of 20 years of ITS research. The resulting analysis indicates a significant shift of the status quo of research in recent years with the advent of novel neural network architectures and the introduction of MOOCs.