Thematic coding of open-ended survey text without a predefined codebook. Claude Opus 4.5 infers each question from the answers rather than the header, clusters 3 to 5 themes, and classifies every participant with no orphans; GPT-5.1 writes the executive layer. Quotes are verified against their source row and deduplicated across themes.
python nlp text-classification text-analysis openai survey-analysis text-clustering qualitative-research survey-research openai-api llm thematic-analysis research-automation anthropic claude-api llm-pipeline qualitative-coding dual-model thematic-coding
-
Updated
Aug 15, 2026 - Python