Several scores and charts are provided to show the percentages of correctness, confidence, and clarity for each test run. This article further explains these charts and scores to help you better understand the information they convey.
Score Thresholds
These are the thresholds used to display the score labels (“OK”, “Good” etc.)
| Label | From | To |
| Great | 95 | 100 |
| Good | 80 | 94 |
| OK | 50 | 79 |
| Bad | 0 | 49 |
Correctness %
The rate at which intents are being classified correctly. A higher score is better.
The Correctness Score Distribution chart displays the balance of correctly predicted intents:
Confidence %
A measure of the confidence scores returned for expected intents. A higher score is better.
The Confidence Score Distribution chart displays the balance for confidence in the predictions:
Clarity %
A measure of the confidence gap between intents. A small gap means that a slight change to the model is more likely to introduce confusion. A higher score is better.
The Clarity Score Distribution chart displays the balance for clarity in the predictions:
Score by Traffic Representation
These charts display the performance of your intents, ordered by its proportion of usage. A high quality model should have the majority of higher traffic intents scoring in the 70+ bracket.Learn more about traffic analysis.
-
Correctness:
-
Confidence:
-
Clarity:
