Natural Language Processing Projects

This repo contains solutions to the projects which were assigned for the Artifical Intelligence II course of the Department of Informatics and Telecommunications at the University of Athens, Greece. The course was specialized in Deep Learning for Natural Language Processing (NLP).

Project 1

The main tasks in this project were to

1. Implement a Ridge Regressor from scratch, along with the batch, stochastic and mini-batch gradient descent algorithms.
2. Develop a sentiment analysis classifier for a given twitter dataset using the scikit-learn python library.
   
   

Project 2

The main tasks in this project were to

1. Solve some theoretical questions.
2. Implement 2 different Feed Forward Neural Network sentiment classifiers for the same dataset used in the first Project. One of them uses the GloVe pre-trained word embeddings.
   
   

Project 3

The main task in this project was to develop and expetiment with different Recurrent Neural Network architectures (i.e. GRU, LSTM) in order to build a classifier for the datased used in the first project. Then, an attention mechanism was added to the RNN in order to improve performance.

Project 4

The main tasks in this project were to

1. Implement a Text Retrieval system using pre-trained Sentence-BERT models.
2. Implement a Question Answering system for the SQuAD2.0 using the BERT model "bert-base-uncased" as a basis.
   
   

Project 5

The task in this project was to write some question in Natural Language that can be used to evaluate a Geographical Question Answering System.

All the programming solutions to the projects were implemented in Google Colabs. The Deep Learning Framework that was used to solve the projects is PyTorch.