Sentiment Analysis on Amazon Fine Food Reviews

This project focuses on performing sentiment analysis on the Amazon Fine Food Reviews dataset using two different approaches: VADER (Valence Aware Dictionary and Sentiment Reasoner) with a Bag of Words approach and the Roberta Pretrained Model from Transformers.

Dataset

The dataset used for this project is the Amazon Fine Food Reviews dataset. It contains reviews of fine foods from Amazon, including a variety of information such as the product ID, reviewer ID, helpfulness rating, review text, and more. The dataset provides a rich source of data for sentiment analysis tasks.

Approaches

VADER (Valence Aware Dictionary and Sentiment Reasoner) - Bag of Words approach

VADER is a lexicon-based approach for sentiment analysis that is specifically tuned for social media texts. It uses a combination of lexical and grammatical heuristics to determine the sentiment of a given text. In this project, we employed the Bag of Words approach, which represents text data as a bag of individual words, disregarding grammar and word order. The VADER sentiment analyzer assigns sentiment scores to each review, indicating the positivity, neutrality, or negativity of the text.

Roberta Pretrained Model from Transformers

RoBERTa is a robustly optimized method for pretraining natural language processing (NLP) models. It builds upon the popular BERT architecture and achieves state-of-the-art performance on various NLP tasks. In this project, we utilized the Roberta pretrained model from the Transformers library. The model was fine-tuned on the Amazon Fine Food Reviews dataset for sentiment analysis. By feeding review texts into the model, we obtained sentiment predictions for each review.

Results

The sentiment analysis results from both approaches were compared and evaluated based on accuracy, precision, recall, and F1-score. The performance of each approach was analyzed to understand their strengths and limitations in sentiment analysis tasks.

Usage

To reproduce the sentiment analysis results, follow these steps:

1. Download the Amazon Fine Food Reviews dataset from Kaggle.
2. Preprocess the dataset by cleaning and preparing the review texts.
3. Implement the VADER approach using a Bag of Words representation and calculate sentiment scores for each review.
4. Fine-tune the Roberta Pretrained Model using the Transformers library on the dataset.
5. Feed the review texts to the fine-tuned Roberta model and obtain sentiment predictions.
6. Evaluate the performance of both approaches by comparing the sentiment analysis results.

Dependencies

Make sure you have the following dependencies installed:

• Python 3.x
• NLTK (Natural Language Toolkit)
• Transformers library

Implementation