README.md

Tweet Sentiment Classification

---

This repo contains Python implementation of Sentiment Analysis of Tweets using Neural Networks with Pytorch.

Overview

• Framework used : Pytorch
• Architectures : LSTM,CNN(2D), and a hybrid of these two
• Dataset : Sentiment140
• Final accuracy : 82.157433 %
• Report

Scripts

• Preprocessing : setup_dataset.py
• Embedding Training : train_embedding.py
• Hybrid Model Training : train_hybrid.py

#Architectures :

CNN based Model

LSTM based Model

Hybrid Model

#Results

Accuracy

• CNN based model : 79.28%
• LSTM based model : 82.5%
• Hybrid Model : 82.15%

Confusion Matrices

• For cnn based model

  

• For LSTM based model

  

• For hybrid model

  

References

---

• Dataset

@ONLINE {Sentiment140,
    author = "Go, Alec and Bhayani, Richa and Huang, Lei",
    title  = "Twitter Sentiment Classification using Distant Supervision",
    year   = "2009",
    url    = "http://help.sentiment140.com/home"
}