Sentiment Analysis NLP Pipeline

This project presents a Sentiment Analysis pipeline built on a dataset of 4000 movie reviews. It encompasses a variety of models including a Naive Bayes Classifier, SGD Logistic Regression, SVMs, and BERT.

File Structure Overview

• entire_pipeline.ipynb:

  • This Jupyter notebook encapsulates the complete pipeline for evaluating various models, excluding BERT.
  • It includes the process for hyperparameter optimization of the SVMs model using grid search.

• bert.ipynb:

  • Focused on the BERT model, this notebook employs both the base uncased and cased versions of BERT for the analysis.

• read_and_preprocess.py:

  • A Python script for reading .txt files and assigning labels based on extracted ratings from filenames.
  • It implements preprocessing steps such as the removal of stopwords and punctuation, and the application of stemming and lemmatization techniques.

• tfidf.py:

  • Features a custom implementation of the Term Frequency-Inverse Document Frequency (TF-IDF) algorithm.

• naive_bayes.py:

  • Contains a custom implementation of the Naive Bayes Classifier.