Speech-Classification

This repository contains models which classify speeches in the Swedish parliament by party affiliation using neural networks. They were developed for my Bachelor thesis in Computer Science. The Hierarchical Attention Network is based on the network in this repository: [https://github.com/JoungheeKim/Pytorch-Hierarchical-Attention-Network/tree/master]

The models

The models can be found in src and can be run using tfidf.py for the tfidf model, main_HAN.py for the HAN model or main.py for all the other models.

Visualizing

The notebook visualize_attention.ipynb provides a visualization of the attention weights in the HAN model.