Skip Navigation
Expand
Detailed information about SmartAssistant AutoTuner
Answer ID 5635   |   Last Review Date 03/18/2019

What do coverage, confidence, implicit data, and explicit data mean in the context of SmartAssistant AutoTuner?

Environment:

Knowledge Foundation
SmartAssistant Autotuner
Oracle B2C Service August 2013 and newer

Resolution:

Explicit Data

Explicit data refers to training data that is explicitly tagged or categorized.  In the case of the SmartAssistant AutoTuner, explicit data means incidents that have best answers selected for them. In comparison to implicit data, explicit data is more valuable because it is more accurate than implicit data.

Explicit training data is used by the SmartAssistant AutoTuner to suggest recommendations and provide information about your current configuration. You can increase the amount of explicit training data scanned by the utility by associating 'Best Answers' to incidents.  For additional details regarding the Best Answers feature, please see Answer ID 6216: Information on the 'Best Answers' button in open incident.

Implicit Data

Implicit data is training data that is generated from uncategorized data based on heuristics.  In the case of SmartAssistant AutoTuner, we examine data from the user_trans table and make inferences about which answers might have resolved a given incident. It is specifically generated with the actions Append as Link and Append as Text.

Implicit data is less accurate than explicit data due to the necessity of making those inferences. However, the data is essentially free in that no human effort is required to generate the data.

Coverage

Coverage concerns the percentage of the population that a sample represents. In the context of SmartAssistant AutoTuner, coverage is a measure of the number of answers that are included in the training data compared to the total number of answers. The higher the percentage of answers included in the training data, the more likely it is that the training is representative of real life usage.

Confidence

Confidence is a measure of how likely the SmartAssistant AutoTuner's suggestion is to actually improve SmartAssistant relevancy. Confidence can be improved by increasing the amount of explicit data available to the AutoTuner. To increase the amount of explicit training data, mark the best answer for incidents that agents have resolved.

For additional information, refer to the 'Adding SmartAssistant Suggested Answers' section in the Online Help User Guide documentation. To access Oracle B2C Service manuals and documentation online, refer to the Documentation for Oracle B2C Service Products.