Brought to you by EarthWeb
IT Library Logo

Click Here!
Click Here!

Search the site:
 
EXPERT SEARCH -----
Programming Languages
Databases
Security
Web Services
Network Services
Middleware
Components
Operating Systems
User Interfaces
Groupware & Collaboration
Content Management
Productivity Applications
Hardware
Fun & Games

EarthWeb Direct EarthWeb Direct Fatbrain Auctions Support Source Answers

EarthWeb sites
Crossnodes
Datamation
Developer.com
DICE
EarthWeb.com
EarthWeb Direct
ERP Hub
Gamelan
GoCertify.com
HTMLGoodies
Intranet Journal
IT Knowledge
IT Library
JavaGoodies
JARS
JavaScripts.com
open source IT
RoadCoders
Y2K Info

Previous Table of Contents Next


6. LESSONS AND RESEARCH ISSUES

Developing an expert system for telecommunications domains typically involves the following stages and each of them has a certain focus:

  1. Selection of an appropriate domain. A successful expert system should tackle the bottleneck of an operation process and have a high impact. Listening to direct user feedback and conducting data analysis are needed to confirm the selection.
  2. Knowledge engineering. Study the domain, collect information, and learn domain-specific knowledge from the experts. Is this a knowledge-rich or data-rich domain? What are the special requirements of this domain? What databases or legacy systems does it need to communicate with?
  3. Selection of AI techniques. Appropriate AI techniques should be able to meet the requirements, and their potential weaknesses have little impact on the specific task. Different techniques can be compared or integrated. Their development and software cost should be taken into account.
  4. Prototype development. A proof-of-concept system should be quickly developed and tried out by users. The preliminary result can be used to refine the prototype. Sometimes, a system can be developed in well-defined phases.
  5. Field trial. A production system must be verified through several field trials. Is its speed efficient? Is its interface user-friendly? Is its operation reliable? At this stage, users do not care which techniques have been used as long as the system does the job.
  6. System transfer and maintenance. An expert system that cannot be maintained or upgraded easily will not live very long because telecommunications technologies and their domain knowledge change constantly. An organization should be identified for the maintenance. A user manual is needed for interactive systems.

During more than a decade of developing expert systems for telecommunications domains, some valuable lessons on using AI techniques have been learned. First, an AI module is often embedded in a large system. It is the core component of the system, but is not everything. Typical telecommunications systems often require database accesses and good graphical user interface. Resource is needed to build these basic modules. Their roles are equivalent important. Second, there are many existing (or legacy) systems that a telecommunications company heavily depends on; an expert system often cannot operate alone. It must talk to these systems. In many situations, one must recognize the fact that an expert system is just a front or rear end of an existing system. Understanding these systems is necessary. Third, not every expert system needs to use the most advanced AI techniques. Simple AI techniques are often sufficient in practice. This reduces the complexity of an expert system. Fourth, expert systems are not a solution for all problems. An expert system approach is needed when it offers the best value among all other alternatives for a specific task. Fifth, to the end-users, an expert system should appear as a useful tool to increase their productivity rather than as a threat to their job security. Expert systems and human operators can often complement each other in terms of their capabilities. Finally, expert systems often need to communicate with each other (within a management layer or across management layers) or with other systems. A common platform and message interface for expert systems is essential for distributed application and system maintenance.

Today's telecommunications domains need AI techniques more than ever. They present a realistic but challenging testbed for AI techniques moving toward real-time, distributed, reliable, and adaptive.

For example, no existing alarm correlation system has the capability to correlate switching alarms, transmission alarms, Signaling System Number 7 alarms, and customer access facility alarms at the same time. The problem could be that there are just too many alarms to correlate in one central deposit. However, if one can limit and customize each monitoring agent to a focused area of problems, it becomes manageable. This requires distributed, personalized monitoring agents.

As another example, telecommunications fraud becomes a major issue in both wireline communication due to wire-tapping or unauthorized use and wireless communication due to customer ID cloning. Detecting fraudulent usage patterns is a nontrivial task because of the vast quantity of data that varies with time and location. This requires real-time data-mining techniques and both short- and long-term trending analyses.

The Internet presents a new opportunity and also poses a threat for the telecommunications industry. It could redefine telecommunications companies as only sellers of big pipes if they do not get into Internet connection and content services. The Internet offers the platform for electronic commerce, voice/data/audio/video communication, and even personal communication by shared virtual reality. Internet content applications include search tools, analysis tools, and personal-filtering tools. For example, GTE provides national electronic yellow pages (http://superpages.gte.net). These tools need AI technology to make them smart, efficient, and adaptive.

Telecommunications tasks are often mission-critical. So far, there is little research on the reliability of AI techniques. When one copy of an expert system dies, how can another copy residing on a different machine recover the inference process and start from a clean state? Some Distributed AI techniques can be useful, but they have not been tried in practice. These are some of the critical tasks that need to be addressed now.


Previous Table of Contents Next

footer nav
Use of this site is subject certain Terms & Conditions.
Copyright (c) 1996-1999 EarthWeb, Inc.. All rights reserved. Reproduction in whole or in part in any form or medium without express written permission of EarthWeb is prohibited. Please read our privacy policy for details.