Fig. 1: Comparison between manual and automated coding process and time efficiency.
From: Evaluating large language models in analysing classroom dialogue
This figure illustrates the process and time efficiency of two methods for analysing classroom dialogue transcripts from math and Chinese classes. The light blue panel describes the automated method using ChatGPT, where text snippets are fed to ChatGPT and the process is repeated nine times with nine snippets, generating results in ~5 min. The light pink panel describes the manual method, where researchers watch videos at 1.5 times the speed, initially read the transcripts, read them again, and then code each turn according to the coding scheme. This manual process takes ~150 min. The automated method is shown to be 30 times faster than the manual method.
