Analyzing Ticket Learning Output

Analyzing Ticket Learning Output

• Go to Intents -> All - And search for the intent you would like to add

• If not found, click on ‘Browse intent library’ and search for the intent

• If found, select the intent and click on ‘Add selected intent’ and activate the intent.

• Go to AI Lens, and click Retrain and Publish and test a few of the newly added phrases to make sure they now trigger the proper intent.

• If phrases trigger the correct intent, then Re-Run the Ticket Learning job to re assess the coverage improvement after adding the intent.

Note: Make sure the intent is not duplicated from the existing. This can be verified against intents that exist in the local and Global.

Top Understood & Not Understood Intents

This is another dimension of showing top intents that are understood and not understood. This view will easily enable users to pick the right priority intent to review.

Top Entities in Understood & Not Understood Tickets

This will show the top entities that are understood from the tickets by validating against the local ontology.

This helps you:

1. Identify the missing fulfillments

2. Attach the fulfillment to the intent manually.

Note: Users need not to do the fulfillment mapping manually, we have a model to recommend the fulfillment automatically by selecting an option under Knowledge Learning -> configurations -> Recommend Fulfillment. However if the user would prefer to search and add manually for the handful of intents they can still follow the below steps.

To identify the Top Understood and Not Understood Entities:

Step 1: Identify top Not Understood Entities which are present at the lower end of the Ticket Learning run output

Step 2: In this case, let’s consider the entity Phone. We want to identify the top Not Understood Intent clusters relating to this entity. To do this go to the magnifying glass icon on the Not Understood Intent tab:

Step 3: Under this icon, you will see a screen like below. On the right you can see the total number of intents pertaining to this entity, and you can click on any of the numbers in the Tickets column to see the cluster of intents.

Step 4: Click on any intent, the screenshot below is the view when you click on the number against Assist Custom: Deskphone

Step 5: An initial glance reveals some broad topics such as:

• Activate Deskphone

• Desk Phone Inquiry

• Desk Phone Issue

Prepare a list of all such possible intents for all top Not Understood Entities, along with the number of tickets associated with them, since this will give us an idea of the coverage we can achieve by working on these intents.

Step 6: Once the intents are ready, we can search for the fulfillments by navigating to the knowledge tab and start searching for Knowledge articles related to these entities. However as mentioned in the Note above, we need not to search for fulfillment manually as we got a model which automates the fulfillment recommendation.Below steps will help you understand the manual capabilities.

You can further hunt for specific issues by trying a "command + f" to narrow down relevant documents.

Top Entities in Understood & Not Understood Tickets

This section will describe the top entities that are understood from the tickets by validating against the local ontology. This helps in two ways

1. 1.Identify the missing fulfillments

2. 2.Attach the fulfillment to the intent manually.

Note: Users need not to do the fulfillment mapping manually, we have a model to recommend the fulfillment automatically by selecting an option under Knowledge Learning -> configurations -> Recommend Fulfillment. However if the user would prefer to search and add manually for the handful of intents they can still follow the below steps.

Step 1: Identify top Not Understood Entities which are present at the lower end of the Ticket Learning run output.

Step 2: In this case, let’s consider the entity Phone. We want to identify the top Not Understood Intent clusters relating to this entity. To do this go to the magnifying glass icon on the Not Understood Intent tab.

Step 3: Under this icon, you will see a screen like below. On the right you can see the total number of intents pertaining to this entity, and you can click on any of the numbers in the Tickets column to see the cluster of intents.

Step 4: Click on any intent, the screenshot below is the view when you click on the number against Assist Custom: Deskphone.

Step 5: An initial glance reveals some broad topics such as:

• Activate Deskphone

• Desk Phone Inquiry

• Desk Phone Issue

Prepare a list of all such possible intents for all top Not Understood Entities, along with the number of tickets associated with them, since this will give us an idea of the coverage we can achieve by working on these intents.Step 6: Once the intents are ready, we can search for the fulfillments by navigating to the knowledge tab and start searching for Knowledge articles related to these entities. However as mentioned in the Note above, we need not to search for fulfillment manually as we got a model which automates the fulfillment recommendation.Below steps will help you understand the manual capabilities.

You can further hunt for specific issues by trying a "command + f" to narrow down relevant documents.