Spoken Dialog System Design
Why is Spoken Dialog System Important
It will further advance the research in the areas such as language understanding,
artificial intelligence and cognitive science.
It has a wide range of applications for efficient human acces to information such as
weather inquiry, airline reservation, customer service, financial data et al.
- There are still some open problems in the areas of language understanding, automatic
speech recognition and text-to-speech synthesis. However, these techniques are mature enough
to build an "intelligent" spoken dialog system for the applications of speech and language
processing in limited domains.
Key Components of Spoken Dialog System
Text-to-speech Synthesizer (TTS): Convert the response of the computer into acoustical signal
and play it back to human.
Natural Language Processor (NLP): Extract the meaning/keywords of the user through the
semantic and syntactic analysis of the input text string from ASR.
Dialog Manager: Connect all other parts of the dialog system together. Plan the action/response
of the dialog system. Control the flow of the dialogue. Get the user information from the database.
Send out the dialog response to TTS.
Database: Store the user information and all other relevant information for the dialog system.
- Automatic Speech Recognizer (ASR): Transcribe the inpur acoustical signal of human
users into a string of text.
System Diagram of a Spoken Dialog System
Speech group at CMU
Lucent Bell Labs
IBM T.J. Watson Center