Enabling Qualitative Research with Small Language Models: A Sliding Window Method for Open Coding on Resource-Constrained Hardware
Rhesa Muhammad Ramadhan, Dedy Sushandoyo · 2025
Open coding plays a foundational role in inductive qualitative research, yet remains difficult to scale in low-resource or field-based settings. Existing tools often depend on cloud-based Large Language Models (LLMs), limiting accessibility due to cost, internet dependency, and data privacy concerns. To address this, we present a method for assisting open coding using locally deployed small language models (SLMs) and a sliding window approach. We evaluated this method on a public dataset of 12 interviews using two distinct SLMs (Qwen 2.5 and Granite 3.1). The results, validated by independent researchers, demonstrate that aligning configuration parameters with the text's discourse structure is key to effectiveness. Crucially, by enabling analysis in situ, this approach resolves a critical methodological delay, allowing researchers to engage with emerging themes during fieldwork. This fosters a more dynamic, iterative research process and delivers a practical framework that positions the researcher as an orchestrator of a more equitable and responsive qualitative inquiry.