Once your model is in production and used to parse sentences, the import rules feature will enable you to define which parsed sentences should be saved or not, according to different criteria such as intent confidence, entity values or other sentence characteristics.
The goal of this feature is to enable you to save sentences coming directly from your users in order to reuse them later on to train your model or test its performance. Rather than saving all conversations and managing sentences individually, which is error prone and time consuming, you can automat this task by saving and categorizing only specific sentences.
You can create and manage import rules in the «Import rules» section.
The home screen of this section shows you all the import rules configured in your project.
Click on a rule to have access to all the sentences that have been saved by this rule and not yet managed in the sentence section.
CREATE A NEW RULE
1. Create a new import rule
Click on the « New import rule » button to create a « New import rule ».
Give it a name and choose the model(s) on which you want to apply this rule. A single rule can be used for multiple models.
Click on next to continue.
2. Configure your rule
Use the filter feature to configure your rule and define which set of sentences will be automatically saved when parsed.
Filters can be used with the following elements:
- Intent confidence
- Nb of labels applied to a sentence
- Text in a sentence
The « entity » filter can be enriched with the following characteristics:
Click on « Next » to continue.
3. Finalize your import rule
Select the label(s) that you want to apply to the imported sentences. This step is optional but we strongly recommend to define at least one label for each rule in order to facilitate their tracking in the sentence section. This will also enable you to create views much easier.
Click on Save to create your rule.
Warning : your import rule will work only if your have selected the « Save sentences » option in the API settings of your model.