Dies ist eine Übersichtsseite mit Metadaten zu dieser wissenschaftlichen Arbeit. Der vollständige Artikel ist beim Verlag verfügbar.
Care for Context: Evaluating Speech-to-Text and Translation Services for 24h Live-In Care Communication
0
Zitationen
3
Autoren
2026
Jahr
Abstract
BACKGROUND: The 24h live-in care sector relies on Eastern European workers and multilingual communication. OBJECTIVES: The aim was to select the best AI-based language services for a 24h care communication app under real deployment constraints. METHODS: A scenario-based quality comparison for transcription, translation, and their cascaded pipeline, and an analysis of the risks and harmfulness for the specific application context were performed. RESULTS: Chirp2 and DeepL showed the best quality. We balanced quality with latency and detected the presence of potentially harmful meaning shifts. CONCLUSION: Domain-grounded evaluation provides a solid base for informed engineering decisions, but considering constraints and risks is necessary for end-to-end safety in deployment.
Ähnliche Arbeiten
Explainable Artificial Intelligence (XAI): Concepts, taxonomies, opportunities and challenges toward responsible AI
2019 · 8.758 Zit.
Stop explaining black box machine learning models for high stakes decisions and use interpretable models instead
2019 · 8.666 Zit.
High-performance medicine: the convergence of human and artificial intelligence
2018 · 8.220 Zit.
BioBERT: a pre-trained biomedical language representation model for biomedical text mining
2019 · 6.896 Zit.
Proceedings of the 19th International Joint Conference on Artificial Intelligence
2005 · 5.781 Zit.