Puhehaku

Puhehaku, or spoken search, is a form of information retrieval where users issue queries by speaking rather than typing. It combines automatic speech recognition with natural language processing to interpret spoken queries and return relevant results. Puhehaku is used in voice-first interfaces, search apps, and multimedia discovery.

Common architectures include capturing audio, automatic speech recognition (ASR) to text, natural language understanding to extract intent and entities, query formulation, search over an index, ranking, and presenting results through synthesized speech or a visual interface.

Applications span mobile devices, virtual assistants, car infotainment systems, smart speakers, accessibility tools, and customer-service IVR.

Techniques emphasize natural-language queries, follow-up clarification, context and history, multilingual support, and, increasingly, on-device processing for privacy. Real-time delivery, robust noise handling, and domain adaptation remain active research areas.

Evaluation uses metrics such as word error rate, search accuracy, mean reciprocal rank, precision and recall, and user satisfaction. Challenges include misrecognition errors that derail retrieval, disfluencies, unusual vocabulary, accents, noisy environments, latency, and privacy concerns.

History and outlook: Spoken search emerged alongside early voice-input systems and matured with mobile devices and cloud-based assistants. Today, puhehaku is embedded in many services as a core interaction paradigm, with ongoing research in accuracy, speed, privacy, and multilingual capability.