Sentiment Analisys on Movie Reviews dataset

Implementation of Sentiment Analysis through Subjectivity Detection and Polarity Classification using a structure seeing involved a Naive Bayes Classifier jointly with a Support Vector Machine (SVM).
The work is inspired by the original paper A Two-Stage Classifier for Sentiment Analysis, Nguyen and Pham.

Install Dependencies

A Conda Environment is provided within the repository to make the dependencies installation easier. To install it please run the following command from the repository root:

    conda env create --file nlu-environment.yml

All dependencies should be included in the environment, however we might still miss out on some dependencies from nltk. A forthright solution to such an issue is to run the following commands after the Conda Environment is set to leave:

    python
    nltk.download()

Run the Code

First off open the Preprocessing Notebook and run all cells. If you prefer to run everything directly from terminal, run the following from reposotory root:

    jupyter nbconvert --to notebook --inplace --execute 01-preprocessing.ipynb

ℹ️ Note: running Preprocessing Notebook first is vital for Main file to work.

Then, to run the main code run the following command from repository root:

    python 02-main.py