The Salience Chart
Showing the impact of each condition on the certainty
Last updated
Showing the impact of each condition on the certainty
Last updated
The Salience Chart is accessed via the and shows a detailed breakdown of the impact of each condition on the certainty, based on the conditions weight and the certainty of the condition.
This view is only available for facts inferred by a rule and can be accessed by selecting the square symbol above the conditions.
Central to the Salience Chart is the certainty of the inferred fact, and the maximum certainty that was possible based on the rule that was executed.
Each condition of the rule will be represented on the Salience Chart, including:
the fact used by that condition (or expression),
the certainty of that fact
the impact this had on the certainty of the inferred fact
The impact of each condition is visually represented to show the actual impact (coloured segment) vs the maximum possible impact (grey segments).
The combined total of all impact percentages will add up to the overall certainty (rounded to the nearest whole number).
In the Salience Chart these will be displayed in grey with 0% certainty and 0% impact. Even though a fact did not exist (i.e. it wasn't injected or answered by a user), when the condition is optional a 'synthetic' fact is inserted with 0% certainty to ensure the rule can pass, but without impacting the certainty.
When creating a rule, the maximum certainty is set to 100% by default, but this can be adjusted to ensure facts can be created with less data and therefore lower certainty.
This means that for a rule with a maximum rule certainty of 60% , even if all conditions are met with facts that are 100% certain, the rule is capped at 60%.
Conditions that do have any impact on the certainty will be displayed to the top-left of the salience chart, under the banner 'Zero salience conditions'.
This is achieved by setting the weight of the condition to zero when configuring the rule.
The impact is based on the , configured in the rule, but also the certainty of the fact used. The higher the certainty of the underlying fact, the higher the impact.
It is possible for facts to be inferred from a rule, even when conditions have not been met (i.e. the user skipped the question), if the conditions has been set as optional. When a rule has conditions that could not be met they will be displayed in the evidence tree fact card with a strikethrough - shown below)
