Speech Interaction Analysis on Collaborative Work at an Elderly Care Facility

Nursing and care are central aspects of healthcare services, and improvements in the quality and efficiency of healthcare processes are important issues. Since these tasks involve both physical actions and information processing, they can be described as what the authors call “action-oriented intellectual services.” There are striking mismatches between such services and current information and communication technology systems, which are generally designed as tools for deskwork. In addition, almost all nursing and care services are realized by the collaborative work of multiple staff members in distributed locations in the field, a situation with which conventional communication media are of limited utility. A smart voice tweet system for nursing and care is proposed to overcome these problems. To realize this technology, one needs to understand how staff communicates to realize collaborative work in healthcare domains. The authors therefore observed bathing assistance, night shift operations, and handover tasks at a private elderly care home for 8 days. The authors collected approximately 400 h of recorded speech, 42,000 transcribed utterances, data from an indoor location-tracking system, and handwritten notes by human observers. The authors also analyzed speech interactions in the bathing assistance task. The authors found that (1) staff members were almost always speaking during tasks, (2) remote communication was rare, (3) about 75% of utterances were spoken to the residents, (4) the intended recipient of utterances was frequently switched, and (5) about 17% of utterances contained personal names. The authors also attempted clustering utterances into what the authors call “passages”, and about 33% of passages contained only one personal name. These results should be applicable in semi-automatic long-term care record taking.