How to test your NLP model

Before testing the functioning of your model, check that all the prerequisites below have already been fulfilled — or click here in case you don’t know what this means.

1. Setting up a provider (Dialogflow, Watson)
2. Addition of at least one intention
3. (Optional) Adding an entity
4. Training and publication of at least one version of the model

After ensuring that all the previous steps have been followed it is time to test your model.

To do this, access the Artificial Intelligence module, choose the Publication submenu, click the Test AI model button and in the side conversation window type the phrase you want to test.

Note that when sending a test phrase Blip will display:

• The intention(s) associated with the phrase;
• What is the reliability (score) associated with each of the intentions.

Your goal now is to test as many phrases as users are supposed to have sent to your company contact. Analyze the possibility of adding new phrases in your intentions or creating new intentions / entities, for examples of phrases not recognized by your model.

• Note: Remember to train and publish the first version of your model before training.
• Note 2: From the second version of the model, new changes must be tested after new training.