Frequently Asked Questions

An Expert System is AI software which replicates the thinking of domain experts in a way that can give expert results. Say, for instance, how to diagnose a fault in your wifi set-up or how to apply for a parking permit. More than a fixed answer, an expert system typically involves a dynamic sequence of targeted questions which hone in on providing a situation specific solution.

The purpose of an expert system is to provide machine reasoned advice at a level comparable to that of a human expert. This capability has two main uses: firstly, to provide professional advice to non-experts and secondly, to provide highly specialised expertise to less qualified or less specialised practitioners.

Experts are people who know things, like how to fix appliances or processes. Meta-experts are experts who know how they know things. This means they can use tools like VisiRule to share that knowledge.

Historical Data often contains patterns which when extracted by Machine Learning algorithms can be used to generate rules. VisiRule can import these rules in the form of PMML decision trees to generate initial VisiRule charts.

Once in VisiRule, these rules are visually exposed and can be refined and tweaked by subject matter experts using their own understanding of how things work.

Backward-chaining rules start with a goal to prove and work backwards, whereas forward chaining rules start with a load of data and looks to use production rules to combine the data into larger composite blocks and go round again. There are some slides from Knut Hinkleman entitled Forward Chaining vs Backward Chaining.

You can also read the article entitled "Forward Chaining vs. Backward Chaining in Artificial Intelligence" written by Parag Radke​

Also, Charles Forgy describes the difference between Forward and Backward chaining.