Creating Synthetic data for Stance Detection Tasks using Large Language models

Alhanouf Alsemairi


Supervised by Fernando Alva Manchego; Moderated by Liam Turner

"Stance" refers to the viewpoint expressed in a piece of text towards a target. For example:

Text: We live in a sad world when wanting equality makes you a troll #SemST Target: Feminist Movement Stance: Favor

This task is usually tackled using supervised classification approaches. However, collecting and annotating datasets by humans can be time-consuming and expensive, thus may lead to a data scarcity problem. This project aims to use Large language models (LLMs) such as ChatGPT and Falcon as potential solutions to address scarcity issues and, therefore, can improve the performance of stance detection models.

This project examines LLMs' abilities in different topics (Feminism Movement, Covid-19 Vaccine) and two languages (Arabic and English).

Final Report (09/09/2023) [Zip Archive]

Publication Form