Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

BEYONDWORDS is All You Need: Agentic Generative AI based Social Media Themes Extractor

Mohammed-Khalil Ghali, Abdelrahman Farrag, Sarah Lam, Daehan Won

TL;DR

BEYONDWORDS presents an integrated framework for scalable thematic analysis of social media using tweet embeddings, autoencoder-based dimensionality reduction, matrix factorization, and agentic Generative AI. The method iteratively refines themes via Chain-of-Thought prompting and a secondary LLM for quality control, demonstrated on autistic community tweets to reveal three core themes: content quality and engagement, advocacy and acceptance, and mental health. The approach achieves robust clustering (three latent themes) and preserves semantic nuance through embedding-based representations, enabling actionable insights for advocacy and decision-making. Its combination of scalable ML techniques and generative AI offers a practical, adaptable tool for analyzing online discourse in diverse communities.

Abstract

Thematic analysis of social media posts provides a major understanding of public discourse, yet traditional methods often struggle to capture the complexity and nuance of unstructured, large-scale text data. This study introduces a novel methodology for thematic analysis that integrates tweet embeddings from pre-trained language models, dimensionality reduction using and matrix factorization, and generative AI to identify and refine latent themes. Our approach clusters compressed tweet representations and employs generative AI to extract and articulate themes through an agentic Chain of Thought (CoT) prompting, with a secondary LLM for quality assurance. This methodology is applied to tweets from the autistic community, a group that increasingly uses social media to discuss their experiences and challenges. By automating the thematic extraction process, the aim is to uncover key insights while maintaining the richness of the original discourse. This autism case study demonstrates the utility of the proposed approach in improving thematic analysis of social media data, offering a scalable and adaptable framework that can be applied to diverse contexts. The results highlight the potential of combining machine learning and Generative AI to enhance the depth and accuracy of theme identification in online communities.

BEYONDWORDS is All You Need: Agentic Generative AI based Social Media Themes Extractor

TL;DR

BEYONDWORDS presents an integrated framework for scalable thematic analysis of social media using tweet embeddings, autoencoder-based dimensionality reduction, matrix factorization, and agentic Generative AI. The method iteratively refines themes via Chain-of-Thought prompting and a secondary LLM for quality control, demonstrated on autistic community tweets to reveal three core themes: content quality and engagement, advocacy and acceptance, and mental health. The approach achieves robust clustering (three latent themes) and preserves semantic nuance through embedding-based representations, enabling actionable insights for advocacy and decision-making. Its combination of scalable ML techniques and generative AI offers a practical, adaptable tool for analyzing online discourse in diverse communities.

Abstract

Thematic analysis of social media posts provides a major understanding of public discourse, yet traditional methods often struggle to capture the complexity and nuance of unstructured, large-scale text data. This study introduces a novel methodology for thematic analysis that integrates tweet embeddings from pre-trained language models, dimensionality reduction using and matrix factorization, and generative AI to identify and refine latent themes. Our approach clusters compressed tweet representations and employs generative AI to extract and articulate themes through an agentic Chain of Thought (CoT) prompting, with a secondary LLM for quality assurance. This methodology is applied to tweets from the autistic community, a group that increasingly uses social media to discuss their experiences and challenges. By automating the thematic extraction process, the aim is to uncover key insights while maintaining the richness of the original discourse. This autism case study demonstrates the utility of the proposed approach in improving thematic analysis of social media data, offering a scalable and adaptable framework that can be applied to diverse contexts. The results highlight the potential of combining machine learning and Generative AI to enhance the depth and accuracy of theme identification in online communities.

Paper Structure

This paper contains 24 sections, 14 equations, 8 figures, 5 tables, 2 algorithms.

Table of Contents

  1. Introduction
  2. Related work
  3. Traditional Thematic Analysis in Social Media Research
  4. Advances in Machine Learning and NLP for Thematic Analysis
  5. Generative AI and Iterative Theme Refinement
  6. Methodology
  7. Text Tokenization and Embeddings Extraction
  8. Embeddings Dimensionality Reduction with autoencoder
  9. Matrix Factorization and Clustering
  10. Generative AI for Themes Extraction
  11. CoT Strategy
  12. Recursive Theme Refinement via LLM Feedback Mechanism
  13. Autism Case Study
  14. Dataset Overview
  15. Dataset Preprocessing
  16. ...and 9 more sections

Figures (8)

  • Figure 1: Framework overview of BEYONDWORDS showing posts embedding, compression, matrix factorization, clustering and generative AI based extraction
  • Figure 2: Overview of the dataset preprocessing used in this research
  • Figure 3: Autoencoder loss plot for a) small, b) medium, and c) large embedding models
  • Figure 4: SVD plot for a) small, b) medium, and c) large embedding models
  • Figure 5: Elbow and k-means themes clusters plot for a) small, b) medium, and c) large embedding models
  • ...and 3 more figures