Date of Award

2018

Degree Type

Thesis

Degree Name

Master of Science in Computer Science

Department

Computer Science and Statistics

First Advisor

Noah Daniels

Abstract

Sentiment analysis within the Natural Language Processing (NLP) field is an active area of research that attempts to classify pieces of text in terms of the opinions expressed. A sub-specialization in this area focuses on classifying or identifying biased text and is growing more important in the era of “fake news.” There are many methods used across researchers so it can be difficult to find a entry point into the field. Not only are there different machine learning methods applied, text embedding techniques have grown in recent years making it difficult to determine the correct avenue to use in research.

This thesis explores different embedding techniques as well as training several machine learning models using sentences from the news annotated using Amazon’s Mechanical Turk (AMT) as either “Unbiased” or “Biased.” Overall, this thesis endeavors to provide an overview of what is currently being done in the field but gathered in one place. The embedding techniques used in this paper focus on predictive models: word2vec, GloVe, and fastText. With each word embedding Support Vector Machines, Neural Networks, Convolutional Neural Networks, and Recurrent Neural Networks. Results show no front-runner in terms of classification accuracy but can still serve as a reference or jumping off point for future research.

Share

COinS