Expired50% confidenceFactExact time
Wake word technology continuously runs a lightweight keyword detection model on-device, activating the full speech recognition and processing pipeline only when a specific word is detected.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
7/2/2026
First Seen
This claim has expired. It is kept for the record and no longer describes the current state.
Valid until: 9/30/2026(expired)
Sources
Hermes Jarvis Deep Dive: The Voice-Driven All-in-One AI Assistant
bilibili小牛AI_XNAI6/7/2026
Related Claims
Unverified传统智能音箱普遍采用意图识别加槽位填充的NLU架构，通过关键词匹配或小型分类模型判断用户意图75% similarUnverified音频指纹技术可以自动识别录音内容，再结合概率模型将其与词曲作品数据库进行关联72% similarUnverified豆包语音输入以字符流方式实时输出识别结果（边说边出字），而非等整句话说完才返回结果，并支持通过系统级快捷键全局唤起69% similarUnverifiedFor detecting implicit repetition at the semantic level, lightweight embedding models such as OpenAI's text-embedding-3-small or BAAI's BGE-small-zh can be used.68% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/46327API
curl https://kongchang.com/api/v1/knowledge/claims/46327MCP
get_claim(id=46327)