Multi-Class_Text_Classification_Deep_Learning_Pytorch

This repository contains various deep learning based multi-class text classification models applied on AG_news Dataset.
It's more of a plug and play type. Designed in a way that you can customize these projects with little or no efforts on custom datasets.

Various Architectures:

• Designing our own Word Embedding + ANN: Creating word embeddings from scratch and Using a Fully connected layer on top of it for classification
• Pre-trained Embedding + Bi-LSTM + FC
• Pre-Trained Fast-Text Embedding + Bi-LSTM / ANN: Fast-Text Embeddings + A configurable architecture to either choose LSTM + FC or FC NN for classification
• Pre-trained Embedding + Bi-LSTM with Attention networks
• Pre-Trained Embedding + Self-Attention
• TextCNN
• Recurrent CNN

To Do:

• Character CNN
• Transformers
• Use precision, recall instead of accuracy for model comparison