AI-generated summary
Genesys Cloud provides comprehensive intent health monitoring and utterance management functionality within its Natural Language Understanding system to evaluate and improve bot training through detailed analysis and modification capabilities. The platform displays an overall health score with actionable suggestions organized by severity, categorizing utterances with labels indicating potential confusion, duplicates, outliers, or excessive length.
The system detects four primary utterance problem types: Potential Confusion identifies utterances too similar to other intents that could cause bot misclassification; Duplicate Utterance flags redundant examples; Outlier Utterance highlights phrases significantly different from intent patterns; and Utterance Too Long indicates training inefficiency. Users can resolve confusion through the Utterance Confusion view, which displays the focused intent alongside related intents containing confusing utterances, enabling side-by-side modification of utterance wording.
The platform enables direct modification of utterances including renaming, rephrasing, removing, and moving between intents via drag-and-drop or the Move Utterance dialog, with health scores automatically recalculated upon saving updates. Changes can be undone only while remaining in the active view; navigation away prevents reversal of modifications. Users can mute suggestions for specific utterances, updating status to "Changes Made" and preserving muted suggestions across user sessions and when exporting bot flows in .i3DigitalBotflow or .i3BotFlow formats.
The interface supports extensive filtering and sorting capabilities. Intents can be filtered by metrics including Very good, Good, Okay, Needs review, Changes made, No data, and New, and sorted by Intent name, Health priority, Utterances count, or Task name. Utterances can be sorted by Priority, Entry Order, or Alphabetical order, and filtered by status categories including Potential Confusion, Duplicate Utterance, Outlier Utterance, Utterance too long, Suggestion(s) Muted, Changes Made, and New. Intent health visibility can be toggled independently through the Natural Language Understanding Intents section; however, the bot continues calculating health metrics regardless of display preference. Intent health scores are calculated individually per bot and may be affected by muted suggestions depending on specific bot configuration.