On the Horizon: Making the Best Use of Free Text Data With Shareable Text Mining Analyses

Abstract

The current sector-wide Enhancement Theme of ‘optimising the use of existing evidence’ encourages the sector to identify what evidence exists, and to explore associated opportunities for best practice. Across the higher education sector, there is a prevalence of free text datasets which are generated through annual surveys and rarely explored across institutions, partly because of the privacy concerns that exist due to the nature of the data. In a recent project exploring secondary analyses of National Student Survey data, the University of Edinburgh also explored text mining approaches to offer fast and repeatable analyses of free text data that can be adopted by other institutions and researchers, without sharing sensitive data. This method has been trialed on institutional level data from the 2016 National Student Survey simultaneously with an in-depth open coding approach to the same data. In this horizons paper, comparisons are drawn between some types of text mining analyses and what can be explored in an open-coding approach, and some recommendations for future use. Alongside this paper is the shareable code for other groups to replicate this approach on their own datasets, to contribute to the optimisation of existing evidence use.