NLP Sentiment Analysis

Project Description

This repository includes two separate projects demonstrating the use of NLP and LLM models.

1. NLP-GPT2: Classifying the tonality of texts using BERT and generating responses based on GPT-2.
2. NLP-LLAMA: Classifying text tones using BERT and generating responses based on LLAMA 3.2.

Dataset used

IMDB Dataset of 50K Movie Reviews

---

NLP-GPT2

Brief Description

This project implements:

1. Tone classification of texts from the IMDB dataset using the BERT model.
2. Generation of responses based on classification results using GPT-2.

Main features:

• Data preprocessing (cleaning emoji, HTML tags, links, etc.).
• Use of custom Dataset for data preparation.
• Pipeline implementation with BERT training and evaluation for tone classification.
• Text generation using GPT-2 based on classification results.
• Metrics: accuracy on training ~91.72%, accuracy on validation ~89.56%.

---

NLP-LLAMA

Brief Description

This project implements:

1. Tone classification of texts from the IMDB dataset using the BERT model.
2. Generation of responses based on classification results using LLAMA 3.2.

Main features:

• Data preprocessing (cleaning emoji, HTML tags, links, etc.).
• Use of custom Dataset for data preparation.
• Pipeline implementation with BERT training and evaluation for tone classification.
• Text generation using LLAMA 3.2 based on classification results.
• Metrics: accuracy on training and validation are similar to NLP-GPT2 results.

---

Installation and startup

Requirements:

• Python >= 3.9
• PyTorch >= 1.12
• Hugging Face Transformers >= 4.30
• CUDA support installed (recommended)