Technology Development Using ATOM
Join our mailing list

Natural Speech Interface Development

Natural Speech Interface Development

The spoken language we usually use everyday is called the natural spoken language. In order to develop a speech interface to which a user can speak naturally, the system needs to be programmed so that it understands and speaks the natural spoken language. Two programs necessary are called Grammar and Dialogue. Developing these two are the core task of implementing a natural speech interface.

ATOM Spoken Dialogue SDK includes
1)Parser:to extract meaning from user's speech
2)Dialogue Manager:to determine how the system should react to a user's speech
3)Natural Language Generator:to convert the meaning of the system into natural language.

These three structures work together with the Grammar Program and the Dialogue Program which are specific to an application, the system can understand what a user says and intends and determine how the system should react.

The architecture of the ATOM Spoken Dialogue System is illustrated below.

  1. Collect and repord potential user speech on an application as data
  2. Speech Recognizer converts user's utterance into text
  3. Develop Grammar File based of the collected data
  4. Based of the developed Grammar, ATOM Parser extracts meaning of user's utterance
  5. Develop Dialogue File which determines action of a dialogue manager
  6. Dialogue Manager decides what action to take (i.e. to ask a question to the user, or to suggest, to reply or to make a specific action according to the user's utterance and the preceding dialogue context)
  7. According to the decision made at (6), Natural Language Generator converts the question into natural language text.
  8. A Text-To-Speech engine converts the result of (7)into spoken language
    (Note:Bold letter describes development tasks)

Grammar and Dialogue Developments

As described above, the development of spoken dialogue systems typically requires to create two main programs. One of them is called Grammar, which makes a computer to understand what user says and what it means. In order for a user to speak naturally and freely, the Grammar needs to cover potentially expected user's utterance as much as possible.
Grammar Development

Another Program is for Dialogue Manaber, which determines how system reacts to user's speech. The goal of a spoken dialogue manager is to make the interaction between user and system natural so that the user can speak freely just as she would with another human.
Dialogue Development

*Agilingua also offers Consulting Services. Please contact us for details.