Using sentiment analysis to predict Amazon ratings

: a comparative study using dictionaries approaches

Abstract

This dissertation delves into the domain of sentiment analysis, a computational approach to detect and extract human sentiments from textual data. With the ever-increasing growth of online textual content, especially in the form of reviews, the need to accurately determine customer sentiment has never been more imperative. To explore the efficacy of lexicon-based sentiment analysis models, this study implements 9 models: VADER, TextBlob, NRC Lexicon, SentiWordNet, Pattern, AFINN, Opinion Lexicon, LabMT, and ANEW. These models are tested on an Amazon reviews dataset, which is uniquely accompanied by a rating system in which the accuracy of the sentiment extraction can be assessed. The study then further delves into a comparative analysis, collecting the performance of these models to discern their strengths, weaknesses, and overall utility.

Date of Award	31 Oct 2023
Original language	English
Awarding Institution	Universidade Católica Portuguesa
Supervisor	Paulo Alves (Supervisor)