Galaxy Communicator Documentation:

Introduction and Overview



History


What you get

The Galaxy Communicator infrastructure is a hub-and-spoke architecture. The distribution contains:

What you don't get

You don't get an end-to-end dialogue system

The Galaxy Communicator distribution is not an end-to-end dialogue system; it provides you with tools for constructing such a system out of a suite of servers. You can obtain the servers you need in a number of ways: MITRE intends in FY02 to assemble an open-source toolkit of Communicator-compliant servers which make up an end-to-end dialogue system, but this effort is in its infancy. At the moment, if you want to use this infrastructure to build an end-to-end dialogue system, you'll need many pieces that aren't provided here, such as audio support, speech recognition, language parsing, dialogue management and language generation.

You don't get run-time semantic standards

The Galaxy Communicator infrastructure provides a sophisticated and general transport layer for connecting servers and Hubs, as well as a message syntax, but does not provide any specifications about the semantics of the messages which travel between the servers and Hubs. That is, there's no standard run-time API for speech recognizers, or audio devices, or parsers. While the MITRE team has explored adapting existing APIs or message sets for use with Communicator-compliant servers, nothing in the infrastructure endorses any of these existing APIs or message sets.

You don't get configuration-time semantic standards

The Galaxy Communicator infrastructure provides no support or specifications for standards for configuring individual servers. For instance, the W3C Voice Browsers group is proposing specifications for speech recognition grammars. Such proposals are completely compatible with the Galaxy Communicator infrastructure, but the infrastructure does not endorse any particular proposal.


Prerequisites: technical background, hardware and software requirements

Technical background

The core Galaxy Communicator infrastructure is written in C, and detailed documentation is provided only for the C libraries, so familiarity with C is fairly important. Since the Galaxy Communicator infrastructure is a distributed infrastructure, some background in distributed processing is preferable (e.g., RPC, CORBA, Java RMI), since distributed processing is a fairly distinct programming paradigm. Object-oriented programming experience is not needed unless you'll be using the Java or Allegro Common Lisp server bindings. Finally (and fairly obviously, since you're already reading this), a reasonable command of English is required for understanding the documentation.

Supported platforms

The current version of the Galaxy Communicator infrastructure is actively supported on Sparc Solaris, Intel Linux and Win32. Consult the installation instructions for more details.

Software requirements

The Galaxy Communicator infrastructure requires the GNU gcc compiler and GNU make. Consult the installation instructions for more details.


Technical overview


A Communicator-compliant dialogue system consists of a process called the Hub, together with a set of servers. Servers and the Hub communicate with each other using named attribute-value structures called frames. These frames form the basis of all structured communication in the Galaxy Communicator infrastructure.

Hub

Almost all communication between Communicator-compliant servers passes through the Hub. The Hub has a number of significant capabilities:

Server libraries

The server libraries provide a number of convenient capabilities for managing the data and communications. Detailed documentation is provided for the C bindings; the documentation for other programming language bindings contains representative examples and equivalence tables.

Please send comments and suggestions to: bugs-darpacomm@linus.mitre.org
Last updated September 28, 2001.

Copyright (c) 1998 - 2001
The MITRE Corporation
ALL RIGHTS RESERVED