LDA Topic Modeling of PERC Papers
This project focuses on using Latent Dirichlet Allocation to thematically analyze the physics education research literature, in the form of the Physics Education Research Conference (PERC) Proceedings in the 2001-2018 period.
The code in this repository was developed by Tor Ole Odden and Alessandro Marin. The project is described in a paper published in Physical Review Physics Education Research:
Tor Ole B. Odden and Alessandro Marin, Marcos D. Caballero, Thematic Analysis of 18 Years of Physics Education Research Conference Proceedings using Natural Language Processing (2020). link.
See the PERC_TopicModeling Notebook. This notebook contains an extract of the methods described in the paper.
To run the main notebook PERC_TopicModeling.ipynb install the required packages:
pip install -r requirements.txt --user
The required packages include Gensim (unsupervised semantic modelling on text), NLTK (Natural Language Tool Kit), LDAVis (interactive topic model visualization), scikit-learn, along with standard data analysis libraries such as pandas, numpy, and matplotlib.
Questions can be directed to Tor Ole Odden