Abstract:

Thematic analysis is the most commonly used form of qualitative analysis used extensively in educational sciences. While the process is straightforward in the sense that a hermeneutic analysis is conducted so as to detect patterns and assign themes emerging from the data acquired, replicability can be challenging. As a result, there is significant debate about what constitutes reliability and rigour in relation to the qualitative coding. Traditional thematic analysis in educational sciences requires the development of a codebook and the recruitment of a research team for intercoder reviewing and code testing. Such a process is often lengthy and infeasible when the number of texts to be analysed increases exponentially. To overcome these limitations, in this work, we use an unsupervised text analysis technique called the Latent Dirichlet Allocation (LDA) to identify distinct abstract topics which are then clustered into potential themes. Our results show that, thematic analysis for educational sciences using the LDA text analysis technique, has prospects in demonstrating rigour and a higher thematic coding reliability and validity, while offering a valid intracoder complementary support to the researcher.

Keywords:

Thematic analysis, topic modelling, Latent Dirichlet Allocation (LDA).