About Classification Training Utility

The Classification Training utility helps train the Virtual Assistant to correctly match customer questions, also known as utterances, to specific intents. This helps the Virtual Assistant to accurately identify the customer’s request and recommend a solution. 

The Virtual Assistant uses intents and utterances contained within a data set known as a model version to manage dialog with the customer. These intents and utterances are tested and verified by the Classification Training Utility before being made active for the Virtual Assistant to use in customer interactions.

Defining Intents, Utterances, and Confidence Levels

The Classification Training utility uses a combination of intents, utterances, and confidence levels to assist the Virtual Assistant in providing customer solutions.

1. Intents: Intents are defined as specific customer requests to perform an action, such as changing an account password, or asking for a specific web page. For example, changing a password or checking an account balance would be considered intents. 

2. Utterances: Utterances are all the different ways in which a customer can ask a question to the Virtual Assistant with the specific intent in mind. For example, How do I change my password? or How do I check my account balance? would be examples of utterances.

   Utterances can also take the form of requests or phrases. For example, I want to change my password or Check account balance would be considered utterances as well, though are not presented specifically in the form of a question. 

3. Confidence Levels: Confidence levels refer to how confident the Virtual Assistant is in its ability to correctly predict the customer’s intent via a particular utterance. Confidence levels fall into three categories: 

   1. High: The intent engine is strongly confident in its ability to predict the customer’s intent.

   2. Medium: The intent engine is moderately confident in its ability to predict the customer’s intent.

   3. Low: The intent engine is not confident in its ability to predict the customer’s intent.

Related Topics

1. Accessing the Utility