QuaLLM: An LLM-based Framework to Extract Quantitative Insights from Online Forums

arXiv:2405.05345v1 Announce Type: new
Abstract: Online discussion forums provide crucial data to understand the concerns of a wide range of real-world communities. However, the typical qualitative and quantitative methods used to analyze those data, such as thematic analysis and topic modeling, are infeasible to scale or require significant human effort to translate outputs to human readable forms. This study introduces QuaLLM, a novel LLM-based framework to analyze and extract quantitative insights from text data on online forums. The framework consists …

abstract analysis analyze arxiv communities concerns cs.cl cs.hc data extract framework however human insights llm modeling quantitative scale topic modeling type world