Hardware Design 1: Initial Sliding Window

Module 3: Current Trends in Expert Systems

Figure: Expert System Applications by Functional Area over Time 3.2 Expert Systems by Industry

A survey of industries conducted between 1984 and 2016 to understand specific industry were expert systems have been applied showed that the popular industries for expert system applications were accounting, financial services, manufacturing and medicine. These are industries that are quick to explore and adopt new technologies such as expert systems. They also typically devote more resources to information technology since they have a greater potential payback. The industry laggards were publishing, real estate, and legal services. It was a little surprising that the transportation industry was also on the low end of the industry applications.

3.3 Problem Domain Trends

The concept of a problem domain describes the general focus or issue of the particular application. Expert systems have to have a very well-defined problem domain in order to function; this is an especially important concept and one of the major determinants of expert system quality. In problem domain delineation, analysis of the problems involve breaking it down into sub-problems. Expert systems are readily applied in easily structured problems.

The chart below shows the problem areas were experts system are applied

Figure: Expert Systems Problem Domains over Time 3.4 Trends in Knowledge Acquisition Techniques

At the outset of expert systems, it was often the case that the domain expert and the knowledge engineer were one and the same. So at that time there was little discussion about the knowledge acquisition (KA) and it attendant bottleneck.

In the past, researchers either borrowed interviewing techniques from the field of psychology, or wrote programs that helped automate and structure the interview process. The manual process was dominantly used in knowledge acquisition, the process involve the structured interview in the knowledge engineering. Tools such as surveys, focused questions, knowledge maps and even the prototype systems themselves to help structure the interview. Since they were popular at the time, some other specific KA techniques such as protocol analysis (verbal, video, and textual) along with KE observation.

After a while, there was much more of a focus on using the appropriate technique for knowledge acquisition Empirical research in the field of expert systems revealed that certain KA techniques are significantly more efficient than others in different problem domains and KA scenarios.

While working on projects some of the categories of KA were updated to include the possibility of computer learning techniques such as ANNs, Bayes Networks and also case-based reasoning being integrated into the project. Computer modelling was expanded to include a variety of computer generated models such as repertory grids, multi-dimensional scaling and semantic networks.

The chart below shows the different knowledge acquisition used in expert systems.

Figure: Expert System KA Techniques over Time

3.5 Knowledge Representation Trends

One of the more difficult determinants to accurately track is that of the particular knowledge representation (KR) schema used in expert system development. While certain knowledge acquisition techniques and expert system platforms may lend themselves to a particular KR schema, in practice there may be many different intermediate KRs used. These could include semantic nets, ontologies, decision trees, cases, neural nets, etc. A target or primary KR schema was identified and additional intermediate KRs were tracked separately when they were included in the case. The different types of KR schema used in expert systems over time are shown in the chart below.

Figure: Expert Systems by KR Schema over Time

4.0 Deploying Expert Systems within the Organization

Determining how and where to use expert systems are based on analyzing the information processing needs of an organization. Identification of places to apply expert system technology is based on finding critical points within an organization where automation of expertise can lead to improvements in operational efficiency. These are some of the reasons that necessitates the adoption and deployment of expert systems in an organization.

 Alleviating "Knowledge Bottlenecks"

"Knowledge bottlenecks" occur when existing expertise cannot be brought to bear on regularly occurring problems that require expertise to solve. That is, "knowledge bottlenecks" happen when the number of experts is too small for the number of problems that need to be solved. Or experts may be geographically distant from the site of the problem.

 Providing a Means for Consistent Decision Making

By deploying expert systems throughout an organization, expertise about narrowly focused problems can be disseminated. This can be used to ensure consistent implementation of policies and procedures.

 Automating Repetitive Tasks That Are Difficult for Humans