Enhancing Automated Text Coding in Online Learning Research: A Systematic Calibration Framework for Large Language Models

Based on the document, the original 1,500 coding entries dataset includes:Dataset Contents:1,500 anonymized text entries from student and instructor comments in online coursesHuman coding results - expert manual coding for all entriesGPT-4o coding outputs - two rounds of automated coding resultsCalibration materials - framework descriptions and calibration notes used for prompt optimizationData Characteristics:Source: 5 online courses (8,459 student comments + 212 instructor comments)Language: Original Chinese text with English translations for framework descriptionsPrivacy: All personal identifiers replaced with \***\ markersScope: Covers 15 different coding frameworks spanning cognitive, emotional, and social dimensionsFormat: Provided as supplementary materials in spreadsheet formatPurpose: This dataset serves as the empirical foundation for evaluating the OLLM-C (Optimized Large Language Model Coding) framework's performance across diverse educational text analysis tasks, enabling transparency and reproducibility of the research findings.