This project aims to classify dialogue acts in conversational data using various model architectures and input features. The goal is to compare the performance of these models and determine which combination of architecture and input features yields the best results.
The code in this project implements the method described in the paper available here.
To install the necessary requirements, run:
pip install -r requirements.txt
To run the model, pass the model name to the main.py script in your terminal:
python main.py --modelname=<model name>
By default, the model uses a GRU encoder GruEncoder. To use the GruSpeakerModel instead, simply pass the name as an argument.
For all other LSTM models, check the notebooks folder for Jupyter notebooks that demonstrate how to use these models.