TERI RESEARCH INC  Providing the Information Advantage...                home.gif (2807 bytes)     

SERVICES       DEFENSE APPLICATIONS      KIOSK-COMMERCIAL       AMERICAN DISABILITIES ACT      back.gif (114 bytes)

Natrual Language System

Product Description

The Natural Language System (NLS) uses a continuous speech, speaker adaptive recognizer, where the system learns to better comprehend the user's unique inflections and pronunciation to optimize understanding. There is no predefined grammar generation required and it has an initial corpora of over 60,000 common words. Unique application semantics, specific words and acronyms are added to the corpora simply by saying and confirming them in a specific domain. Unconstrained speech recognition is appropriate for applications where either free-text input is required or possibly where users are unfamiliar with the input jargon. With NLS the user has unlimited ways of conveying intent without sticking to a script.

For example, if a user is trying to make airline reservations, the date, time, cities and airline are all unique variables that can be uttered in any grammatical combination. This capability would also be appropriate for web browsing or search-engine speech input as well. The NLS application suite also includes an understanding paradigm, to infer the intent of the speaker, within the target application domain. Interactive speech dialogue is supported to provide clarification and to guide the conversation.

Operational Features

The advantage of the NLS product is its unique ability to develop a natural language understanding of the users spoken intent, without grammatical syntax boundaries. This is a natural and intuitive speech enabling interface that offers truly hands-free and eyes-free operation of computer and communication devices in mobile and public environments.  There are two modes of operation available, when using the NLS. The first is when the application requires specific input information to substitute for mouse, keyboard or touch screen, even though the utterance does not follow a pre-defined syntax.   The second NLS mode of operation is unconstrained free-text. Here the user is able to dictate any desired text, edit and forward it, using spoken commands. Real time text generation from spoken language input offers the opportunity for the user to naturally free-form spoken requests to access or search web information.

A domain learning tool is used in the NLS to define the valid command and control activities. It identifies the components of grammatical speech to be recognized from the unconstrained input text, and provides a contextual interpretation of the intent of the speaker. Once input speech has been detected by the recognizer, it is sent as a parsed grammatical phrase to the Dialogue Manager. Then the grammatical component data is processed by TRI's natural language understanding algorithms to determine the speakers intent. Once an intent has been established, that matches a domain pre-defined action or activity, a token or virtual mouse click is generated and sent to the application for action.

The NLS also uses a number of flexible dialogue initiatives and speech synthesis to converse with the user, and to alert the user of particular situations or options. Also, real-time dialogue is used to provide help, ask for clarification of intent, annunciate a received message or alert, or generate information feedback based upon the application domain.

NLS Packaged Components

Runtime Product Price $999.           Development Product Price $2599.