Personality prediction based on large language models

• Main Author: Wee, Jewel Xin Yu
• Other Authors: Erik Cambria
• Format: Final Year Project
• Language: English
• Published: Nanyang Technological University 2024
• Subjects: Computer and Information Science; Personality prediction; Myers-Briggs type indicator; Large language models; Sentiment analysis
• Online Access: https://hdl.handle.net/10356/175297
• Institution: Nanyang Technological University

Automated personality prediction from text, with the help of Machine Learning (ML) and Deep Learning (DL) Natural Language Processing (NLP) techniques, has been widely researched in recent years. It offers a solution in tackling issues faced by self-assessments, particularly response bias. It may also enhance the areas of marketing, advertising and customer service. However, with the current surge in the use of generative Artificial Intelligence (AI), there is significant opportunity to investigate the novel approach of leveraging Large Language Models (LLMs). Thus, this study aims to bridge the gap in the lack of use of LLMs for personality prediction. Text data from social media sites such as Reddit and PersonalityCafe will be used to infer Myers-Briggs Type Indicator® (MBTI) Types. Additionally, sentiment analysis will be performed to discover linguistic and emotional patterns related to personality.