Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

LOGOS: LLM-driven End-to-End Grounded Theory Development and Schema Induction for Qualitative Research

Xinyu Pi, Qisen Yang, Chuong Nguyen

TL;DR

LOGOS is introduced, a novel, end-to-end framework that fully automates the grounded theory workflow, transforming raw text into a structured, hierarchical theory, and demonstrates a potential to democratize and scale qualitative research without sacrificing theoretical nuance.

Abstract

Grounded theory offers deep insights from qualitative data, but its reliance on expert-intensive manual coding presents a major scalability bottleneck. Existing computational tools either fail on full automation or lack flexible schema construction. We introduce LOGOS, a novel, end-to-end framework that fully automates the grounded theory workflow, transforming raw text into a structured, hierarchical theory. LOGOS integrates LLM-driven coding, semantic clustering, graph reasoning, and a novel iterative refinement process to build highly reusable codebooks. To ensure fair comparison, we also introduce a principled 5-dimensional metric and a train-test split protocol for standardized, unbiased evaluation. Across five diverse corpora, LOGOS consistently outperforms strong baselines and achieves a remarkable average $80.4\%$ alignment with an expert-developed schema on complex datasets. LOGOS demonstrates a potential to democratize and scale qualitative research without sacrificing theoretical nuance.

LOGOS: LLM-driven End-to-End Grounded Theory Development and Schema Induction for Qualitative Research

TL;DR

LOGOS is introduced, a novel, end-to-end framework that fully automates the grounded theory workflow, transforming raw text into a structured, hierarchical theory, and demonstrates a potential to democratize and scale qualitative research without sacrificing theoretical nuance.

Abstract

Grounded theory offers deep insights from qualitative data, but its reliance on expert-intensive manual coding presents a major scalability bottleneck. Existing computational tools either fail on full automation or lack flexible schema construction. We introduce LOGOS, a novel, end-to-end framework that fully automates the grounded theory workflow, transforming raw text into a structured, hierarchical theory. LOGOS integrates LLM-driven coding, semantic clustering, graph reasoning, and a novel iterative refinement process to build highly reusable codebooks. To ensure fair comparison, we also introduce a principled 5-dimensional metric and a train-test split protocol for standardized, unbiased evaluation. Across five diverse corpora, LOGOS consistently outperforms strong baselines and achieves a remarkable average alignment with an expert-developed schema on complex datasets. LOGOS demonstrates a potential to democratize and scale qualitative research without sacrificing theoretical nuance.

Paper Structure

This paper contains 62 sections, 6 equations, 5 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Rationale and Conceptualization
  3. Problem Formulation.
  4. Method
  5. The LOGOS Framework
  6. Step 1 - Chunking & Open Coding.
  7. Step 2 - Embedding & Clustering.
  8. Step 3 - High-level Code Generation
  9. Step 4 - Code Pair Relation Classification.
  10. Step 5 - Graph Construction.
  11. Step 6 - Codebook Clean-Up.
  12. Iterative Codebook Refinement Mechanism
  13. Statistical Metrics for Codebook Quality
  14. Reusability.
  15. Descriptive Fitness.
  16. ...and 47 more sections

Figures (5)

  • Figure 1: The codebook fitting and prediction mechanism in the LOGOS workflow overview. Retrieval based refinement during training shares the same workflow as retrieval based coding.
  • Figure 2: Statistical metrics breakdown.
  • Figure 3: Breakdown of five distributional metrics dimensions and codebook size variations.
  • Figure 4: On Math Failure dataset, GraphRAG and LightRAG fails to capture failure errors because no relevant information is extracted. In contrast, LOGOS successfully deliver a sensemaking answer.
  • Figure 5: Code match heatmap.