This thesis designs, implements, and evaluates an in-platform workflow that improves the searchability and accessibility of dialect-rich cultural heritage audio by combining Automatic Speech Recognition (ASR), volunteer transcript correction, and content annotation within Isof’s Folke crowdsourcing platform. Using the Action Design Research approach and a sociomaterial lens, the project proceeded in two cycles over nine months. Cycle 1 developed a content annotation workflow, while Cycle 2 added a per-utterance ASR transcript reading and correction workflow. The evaluation drew on surveys and interviews with experienced crowdsourcing contributors, two reference group workshops with Isof staff, observational notes, and internal documents.

Findings indicate that even imperfect ASR drafts lower the start-up cost of working with long, noisy recordings and help focus human effort. Micro-tasks with instant confirmation reduce cognitive load; pairing controlled vocabulary with adjacent free-text notes preserves interoperability while capturing contextual nuance. Visible task status and contribution counters support motivation, and institutional alignment supports long-term sustainability. Main contributions: (1) a working high-fidelity prototype integrating ASR transcript correction with content annotation in Folke; and (2) five transferable design principles for cultural heritage organisations.