Teacher's classroom discourse analysis

This project presents the R code used to prepare and analyse data coming from a lesson recording.

The data is composed of the transcript of a teacher’s discourse during an informatics lesson, with some comments.

It is part of a PhD project in educational sciences.

This file presents the R code for the analysis of a lesson where one single teacher is working and talking.

For data protection reasons, the data is not available in this repository. We only provide R code to give an idea of the process setup to analyse the data.

Apart from the RStudio project file itself (lesson_discourse_1t.Rproj), the main file is a Quarto document: class_id_lesson_id.qmd.

From this file, and as in our context we use this project multiple times to analyse different lessons, we centralise most of the code and import several .R files.

• 01_libraries.R imports all the packages needed in the project
• 01bis_import_initial_variables.R imports the variables provided by another analysis (dataframes called classes, lessons, discourses, teachers that centralise all the variables related to the classes, the lessons, the discourses and the teachers
• 02_import.R imports transcripts and comments as .csv files coming from the Trint platform
• 03_end_of_intro.R cleans the data, merges discourse and comments, creates the main Quanteda corpus and tokensobjects and computes textual statistics
• 04_splitting_lemmatisation_tokenisation_dfm_creation.R splits the corpus into segments for the clustering, lemmatises with Spacy, adds as docvars() all the variables to the corpus and creates a data-feature matrix (Quanteda dfm object)
• 05_plot_frequencies.R plots the most frequent words in the lexicon
• 06_reinert_clustering.R computes the clustering with rainette package, identifies the biggest clusters and plots the evolution of clusters over time
• 07_alt_creation_and_ca.R creates an aggregated lexical table (ALT), computes a Correspondance Analysis (CA) using CA() from FactoMineR package, plots it and prepares tables showing the elements most related to the first two axes of the CA
• 08_export.R exports the Quanteda corpus and tokens objects for the global analysis grouping all the lessons, as well as all the variables computed as dataframes