Rukshan Edirisinghe

Portfolio detail

Objectives:

• Analyze Review Sentiment
• Compare Model Performance
• Gain Sentiment Insights

Libraries Used:

This Python program uses various libraries for tasks such as sentiment analysis, data manipulation, and visualization:

• Pandas
• Numpy
• Matplotlib.pyplot
• Seaborn
• NLTK (Natural Language Toolkit)
• tqdm
• transformers
• torch

Summary:

The project is a walkthrough of a sentiment analysis project using Python. It covers both traditional approaches using NLTK's Vader model and more advanced approaches using Hugging Face's pre-trained RoBERTa model. The dataset used consists of Amazon Fine Food reviews, containing text reviews and corresponding star ratings ranging from one to five.

• Data Loading and Exploration: The dataset containing reviews and star ratings is loaded using Pandas, and basic exploratory data analysis (EDA) is performed.
• Text Processing and Sentiment Analysis with NLTK: The NLTK library is used to tokenize sentences, tag parts of speech, chunk entities, and perform sentiment analysis using the Vader model.
• Sentiment Analysis with RoBERTa: Hugging Face's RoBERTa model is utilized for sentiment analysis, leveraging its advanced capabilities to capture contextual information in text.
• Comparison and Visualization: The results from both models are compared and visualized using Seaborn's pair plot and bar plots to assess their performance and identify any discrepancies or patterns.
• Example Analysis: Specific examples are examined where the sentiment analysis models exhibit unexpected behavior, highlighting the strengths and limitations of each approach.

Overall, the project demonstrates the application of different techniques and libraries for sentiment analysis, showcasing the evolution from traditional approaches to state-of-the-art transformer models.