Book discussions message classification

Project contains the implementation of several models and feature extractors for message classification.

Report

Report is available here.

Setup

1. Install the required packages

pip install -r requirements.txt

Running the experiments

A simplified version of the experiments (using a single dataset split instead of cross validation) can be run using the following command. The script should finish in a minute or two.

python classification.py

To run the full experiment used to produce the plots and tables in the report, use the script with the --full flag.

python classification.py --full

The notebook cross_validation.ipynb contains the same script in notebook form and is a good starting point for playing around with custom models and feature sets.