Sentiment and Interest Detection in Social Media using GPT-based Large Language Models
Md Abdullah Al Asad, Hasan Md Imran, Md Alamin, Tareque Abu Abdullah, Suriya Islam Chowdhury · 2023
In the ever-expanding realm of social media, deciphering sentiments and identifying relevant topics from various textual content is challenging. This paper comprehensively investigates sentiment analysis and interest/topic detection using cutting-edge language models, ChatGPT3.5 and gpt4All. The methodology encompasses data collection, meticulous text pre-processing, innovative prompt design, and the exploration of zero-shot, one-shot, and few-shot learning techniques. We unveil the nuances of model performance in sentiment analysis and interest/topic detection through a detailed comparative analysis. Our findings highlight the power of ChatGPT3.5 in achieving a substantial accuracy enhancement in sentiment analysis compared to gpt4All. Moreover, we delve into the intricacies of interest detection, demonstrating the complexities of linguistic structures and model biases. We offer a holistic view of social entities’ preferences by categorizing topics into distinct domains. A web portal developed using Google’s Flutter SDK facilitates the visualization of user-friendly sentiment and interest outcomes. This research contributes to the understanding of sentiment analysis and interest detection and underscores the evolving capabilities of AI and NLP in navigating the dynamic landscape of social media content.