FAQ: Phrases Identified / Highlighted for Sentiment Detection

1. How does the system identify phrases for sentiment detection?

        Our sentiment detection model is trained using a set of standardized phrases and language patterns that are categorized as Positive, Negative, or other defined signals.

These phrases help the ML model understand tone, intent, and customer sentiment more accurately across cases and communications.

2. Can customers access the list of standard phrases used for sentiment detection?

        No. The predefined phrase library and internal signal mapping are part of our proprietary ML framework and are not shared externally. This ensures model integrity, consistency, and continuous improvement.

3. What if a phrase is incorrectly classified or not highlighted properly?

        If you notice that a phrase should be identified differently (for example, marked as Positive or Negative but is not), you can label or highlight it directly in the UI (where applicable).

Once submitted:

• The feedback is captured and shared with our ML team.
• The team evaluates the request.
• Based on impact, frequency, and priority, the enhancement is considered for model training updates.

4. How often are sentiment model updates implemented?

        Model improvements and refinements are typically reviewed and implemented on a quarterly basis.

Approved enhancements and prioritized phrase updates are incorporated into the ML training cycle and rolled out as part of these updates.