Preparing Stance Detection: Feature Extraction Methods and Their Performance Used for Feature-Based Machine Learning Algorithms

Preparing Stance Detection: Feature Extraction Methods and Their Performance Used for Feature-Based Machine Learning Algorithms

Vass, Kristóf (TU Delft Electrical Engineering, Mathematics and Computer Science; TU Delft Intelligent Systems)

Murukannaiah, P.K. (mentor)
Marroquim, Ricardo (graduation committee)

Delft University of Technology

Computer Science and Engineering

CSE3000 Research Project

2021-07-01

Stance detection is a Natural Language Processing task that aims to detect the stance (support, agreement, or disagreement) of a piece of text towards some target. In this paper, we aim to find the best performing pair of feature extraction method and feature-based machine learning algorithm. By doing so, an explainable method can be found to show how to solve stance detection problems. After researching the most common techniques, twenty different combinations are evaluated. We have found that the best performing pair is Word N-gram used with Logistic Regression, which achieves an F-score of 0.599 and an accuracy of 0.66.

Stance Detection
Machine Learning
Natural Language Processing

http://resolver.tudelft.nl/uuid:82ada98d-f99e-4f05-8d9d-9d0831c68c73

Student theses

bachelor thesis

© 2021 Kristóf Vass