Automatic topic detection on Chinese essays: a technology enhanced approach to facilitating formative use of summative assessment

Abstract

Essay writing is an important form of assessment. It is typically used as summative assessment and may not have some of the major benefits of formative assessment. Although there are considerations of formative use of summative assessment, such use remains limited in practice. It is because of the difficulty in presenting the results and the large amount of effort required. In this paper, we propose to use an automatic topic detection method, named hierarchical latent tree analysis, for analysing students' essays. The method can identify topics in the essays and organise those topics in a hierarchy with multiple levels of granularity. It can possibly facilitate formative use of summative assessment by addressing the main reasons prohibiting such use. We present the empirical results of the method using the 54 Chinese essays written by students in an undergraduate course. We further discuss and demonstrate how the method can facilitate formative use of the essays.