Latent Dirichlet Allocation

Code for running robust and repeatable LDA experiments. Use the -h flag to view CLI parameters for any script.

Files and Folders

• lda: Files related to the training and analysis of LDA topic models
• dlda: Files related to the training and analysis of dynamic topic models (using gensim's ldaseq implementation)
• list_common_words.py: Takes an experiment config file as a command line argument and runs all specified preprocessing before listing the top 50 words in the dataset which will be used in that experiment
  • See lda or dlda READMEs for the structure of an experiment JSON file
• plot_data_quants.py: Driver function to use a TextParser to make plots of the quantities of data in time frames (especially useful for deciding time intervals for a dynamic topic model)

Dependencies

Install our ogm package and its dependencies.