Jump to: navigation, search

Applying the Ratings

This topic describes part of the functionality of Genesys Content Analyzer.

Ratings of models, described in Testing Models, have several possible uses.

Assessing Overall Quality

To assess the overall quality of a model, look at the ratings on the Average Results Subtab and the Correct in Top N Subtab.

Identifying Confusing Categories

To identify categories that may be too similar and/or have insufficient feedback, use the ratings on the Category Confusion Subtab. Consider modifying any pair of categories that has a confusion rating over 0.2.

Using in Routing

To decide where to set the Confidence level in an IRD Classify object (see “eServices Objects” in the “Interaction Routing Designer” chapter of the Universal Routing 8.1 Reference Manual, use the ratings on the Average Results Subtab and the Results by Category Subtab. The way that you use these ratings differs according to the task that you are performing at this step of the strategy:

  • If you are using the classification to choose an autoresponse or acknowledgment: Set a relatively high Confidence value, one at which Precision is 85 or higher and Recall is 5 or higher.
  • If you are using the classification to choose standard responses as suggestions to the agent: Set a Confidence value of 1. Even a very low Precision rate (for example, 15) is safe because an agent will make the final decision on whether to use the standard response. Also the lower the Confidence value, the more categories are returned, and:
    • The higher the probability that the correct category is among them.
    • The more categories the agent can provide feedback on.
  • If you are using the classification to determine where the interaction goes in the next step of the strategy: Set the Confidence value at the point where Precision is approximately equal to Recall.


This page was last modified on December 19, 2013, at 16:22.

Feedback

Comment on this article:

blog comments powered by Disqus