NLU
A subfield of natural language processing focused on enabling computers to comprehend the semantics and intent of human language. Core tasks include intent recognition and entity extraction. NLU serves as a fundamental intermediate layer in dialogue systems and voice assistants.
Core Facts
Traditional voice AI systems adopt a cascaded architecture that splits voice processing into independent modules such as ASR, NLU, dialogue management, and TTS, with information loss and latency accumulation between each module.
传统语音AI普遍采用级联式架构，先用ASR将语音转文字，再交给语言模型处理，最后用TTS合成输出
Traditional voice control systems are based on a pipeline architecture of ASR (Automatic Speech Recognition) + NLU (Natural Language Understanding) and rely on predefined instruction templates and keyword dictionaries.
Timeline (last 90 days)
The chaining of three layers of capability—speech recognition (ASR), natural language understanding (NLU), and scheduling planning—determines whether voice commands can be executed accurately, and an error in any layer leads to misoperation.
The core architecture of Agent Assist products consists of a bottom-layer STT engine for speech-to-text conversion, a middle-layer NLU module for intent recognition, sentiment detection, and entity extraction, and a top-layer decision engine for pushing knowledge cards and script suggestions.
All Facts (8)
Traditional voice AI systems adopt a cascaded architecture that splits voice processing into independent modules such as ASR, NLU, dialogue management, and TTS, with information loss and latency accumulation between each module.
80%Verified传统语音AI普遍采用级联式架构，先用ASR将语音转文字，再交给语言模型处理，最后用TTS合成输出
75%VerifiedTraditional voice control systems are based on a pipeline architecture of ASR (Automatic Speech Recognition) + NLU (Natural Language Understanding) and rely on predefined instruction templates and keyword dictionaries.
75%Unverified智能客服的发展经历了「关键词匹配→规则引擎→机器学习分类→深度学习NLU→大语言模型」的演进路径
85%Unverified早期NLU依赖规则匹配和词袋模型，准确率低且泛化能力差
85%UnverifiedThe chaining of three layers of capability—speech recognition (ASR), natural language understanding (NLU), and scheduling planning—determines whether voice commands can be executed accurately, and an error in any layer leads to misoperation.
50%UnverifiedThe core architecture of Agent Assist products consists of a bottom-layer STT engine for speech-to-text conversion, a middle-layer NLU module for intent recognition, sentiment detection, and entity extraction, and a top-layer decision engine for pushing knowledge cards and script suggestions.
50%Unverified传统语音助手长期面临智能停留在功能调用层面、缺乏真正语义理解与情境推理能力的批评
50%