<![CDATA[Hospitals today face growing pressure to evaluate service quality in ways that go beyond what conventional surveys can offer. Patient satisfaction questionnaires remain widely used, yet their dependence on retrospective recall means that problems are often identified long after they occur, if at all. This study set out to explore whether patient conversation data, captured via passive audio recording and processed by speech recognition, could serve as a near-real-time source of service quality indicators in a hospital setting. An exploratory observational approach was taken at a private type C hospital in the Special Region of Yogyakarta. Recordings were gathered from public service areas. After processing and cleaning, the dataset contained 47,223 textual units, with 7,368 unique words and 22,329 meaningful phrases. The analysis drew on frequency counting, thematic categorisation, sentiment orientation, and linguistic diversity mapping. Waiting time and queuing issues dominated the data, with 'queue' appearing 317 times and 'waiting' 276 times. BPJS-related topics appeared 155 times, referral problems 114 times. Negative expressions outnumbered positive ones. Conversation volume peaked at 07.00-09.00 (40% of all recorded activity). The findings suggest that routinely recorded patient conversations, when processed systematically, may serve as a practical complement to conventional satisfaction surveys, showing promise for supporting more timely and evidence-grounded decisions in hospital service management.]]>
Copyrights © 2026
