Engage Digital | Configuring AI filtering

When you are ready to configure AI filtering, follow these steps:

Adding an AI filtering configuration

To add an AI filtering configuration, follow these steps:

1. Navigate to Routing > AI engine via the left-hand navigation bar.

2. Click Add then AI Filtering.

3. Enter the value you want for Minimum precision.

4. Enter a Start date for messages to be processed.

5. Select a Status for the configuration.

6. Select the channels the configuration applies to from the Channels dropdown menu.

7. Check Auto-compute best params. (optional)

8. Check Use co-occurrences. (optional)

9. Enter a value for Min words frequence and Max words in model.

10. Click Save.

Iterating on model results

Once your model has completed a learning cycle, you can review results and edit model parameters if needed:

1. Navigate to Routing > AI engine via the left-hand navigation bar.

2. Click Edit next to the AI filtering configuration entry.

3. Review the precision coverage images to see precision results of your model.

4. Review the training.txt file to see detailed statistics for your model.

5. In the model configuration settings, readjust any of the settings as needed.

6. Enter any Keywords needed by the model for each channel.

7. Click Save.

AI filtering configuration options

When creating an AI filtering configuration, you’ll have access to a variety of configuration options, which can be found in the Create an automatic categorization configuration page. These settings include basic parameters for the configuration itself, as well as parameters that define the model. Let’s review the configuration options available when creating an AI filtering configuration:

Minimum precision: Specifies a level of precision below which AI engine will not perform filtering.
Start date: Specifies a date for the messages to start being processed by the AI engine.
Status: Specifies the operating mode of the AI filtering configuration.
- Inactive: Disables the AI filtering configuration.
- Synchronize model, no autocategorization: Runs in a simulation mode to determine the reliability of the model before it is launched.
- Synchronize model and autocategorize: Runs in the typical operating mode.
- Autocategorize, no model synchronization: Categorizes messages without using the learning function.
Channels: Specifies the channels from which the AI engine will categorize and build its model
Auto-compute best params: Instructs the model to automatically calculate the best model settings.
Use co-occurrences: Instructs the model to analyze commonly associated words as a single term.
Min words frequence: Specifies the number of times a word must appear in message content to be included in the model.
Max words in model: Specifies the maximum number of words to be included in the model.
Keywords: Specifies keywords needed by the model for each channel. (available after model learning)