Train and publish NLU model
1E ITSM Connect uses ServiceNow's Natural Language Understanding(NLU) to comprehend word meanings, recognize word contexts, and infer user or system actions. 1E ITSM Connect specifically uses this NLU engine to perform recommended actions feature. The 1E ITSM Connect application has its own NLU model and can be customized based on the requirement. In the customer environment, this model is trained and published automatically using fixed scripts. However, to manually configure it, follow the below given steps.
Login to your ServiceNow instance with admin credentials.
Navigate to NLU Workbench → Models in the application navigator. The page NLU Workbench opens up with the list of All existing models on the right side of the window panel. Look for 1E Recommended Actions NLU (English) model in the list as shown in the picture below.
Now click on the model 1E Recommended Actions NLU (English) to open the record. This opens up the model details page as shown in the picture below.
Click on the Intents card or click on View phase button given on the Build and train your model section to open into Build and train your model page as shown below. This page has a list of all the built-up intents and their details in the selected model. Now click on the Train model button given on the right side.
A short summary panel appears on the right side of the window, briefing about the content changes that will take place.
Click on the Train button to initiate the process of training the model. This may take a few minutes to complete the process. When the model is trained successfully, it displays a success message to the user as shown below.
Now go back to the 1E Recommended Actions NLU model details page and click on View phase button under Test and publish your model section as shown in the picture below.
This opens up Test and publish your model page as shown in the picture below. On this page you can click on Run new Test button, if you want to add more utterances to your intents.
This may take a few minutes to complete the process. When the model is published successfully, it displays a success message to the user as shown below.
Once the model is trained and published successfully, it is now required to map instructions with intents. For mapping instructions to intents, refer Mapping intents with 1E instructions.