Textual and Contextual Factors in the Performance of Sentiment Classifiers

This repository contains the notebooks that were used for doing sentiment analysis on the IMDB dataset.The different classification models that were used are SVM, Naive Bayes, LSTM, Logistic Regression and BERT. All the notebooks are placed in the Notebooks folder found in this repository. Each of the notebook are named according to the model used in the notebook. The dissertation_main file contains the SVM, Naive Bayes, Logistic Regression and LSTM models. The other experiments like bigram and trigram and also first part, second part and third part accuracies file contains the experiment of splitting sentence into 3 parts which has also been included in the repository. The Splitting_sentence_into_three_parts file is the notebook were the review is split into three according to number of words in the review.These tables are then later used for finding the accuracies.