This repository contains three reports prepared as part of a university course on Natural Language Processing. Each report focuses on a different methodology for sentiment classification on Twitter data. The code is not published in this repository; only the reports are included.
-
Project 1: TF-IDF + Logistic Regression
A classical machine learning baseline using TF-IDF features and a logistic regression classifier.
View Report -
Project 2: Word2Vec + Deep Neural Network
A deep learning approach leveraging Word2Vec embeddings combined with a custom neural network implemented in PyTorch.
View Report -
Project 3: BERT Fine-tuning
A transformer-based model using (Distil)BERT fine-tuned for sentiment analysis, achieving the highest performance among the three approaches.
View Report
The repository serves as a showcase of the progression from traditional machine learning methods to modern deep learning and transformer-based models for sentiment analysis. It highlights the theory put into practice, evaluation metrics, and lessons learned at each stage.
- The reports are provided as submitted for academic evaluation.
- Datasets and source code are not included.