Correcting training data in an email bot
When you want to improve the ability of Pega Email Bot™ to detect topics, language, and entities, you can review and correct the training data in the system. By correcting the training data and rebuilding the text analytics model, you improve the artificial intelligence of the email bot and teach it to more accurately detect the desired information in emails. The system can then suggest the right business case or email response, based on the detected information.For example, you can correct training data related to a car insurance request for an email bot, so that later the system correctly identifies a car insurance topic and entities related to a car make and model.
When you enable the recording of training data and the email bot receives an email, the system saves the email as a training record. (You can also manually create a training record that contains sample information for an email from a user). Once you correct the data detected in the training record, you mark the record as reviewed. The system then uses the reviewed training data to rebuild the improved text analytics model.
In the navigation pane of App Studio, click Channels.
In the Current channel interfaces section, click the icon that represents your existing Email channel.
In the Email channel, click the Training data tab.
If you configure multiple languages for the email bot, to filter data records by a language, in the Language list, select a language.To display data records only detected in the German language, select German.
In the list of training records, select a data record.The Review training data section displays the detected entities, and the NLP analysis section displays the detected language, topic, and entity types for the training data record.
Correct the data for the training record:
Choices Actions Modify text
- In the Review training data pane, click the More icon, and then click Edit.
- In the Update text window, edit the text for the training record, and then click OK.
Update the topic to be detected
- In the Topic field in the NLP analysis section, press the Down arrow key.
- Select a more appropriate topic for the training record.
For example, to correct a training record so that its intent relates to car insurance, select Car insurance.
Update the language to be detected In the NLP analysis section, in the Language list, select a language. Add new entities
- In the Review training data section, in the data record content, highlight and right-click the text that you want to map to the new entity, and then click New entity.
- In the Create new entity window, in the Entity name field, enter a name for the entity, and then click Submit.
For example, to create an entity for a car make, highlight the word Ford and enter Car Make as the new entity.
Update existing entities In the Review training data section, in the data record content, highlight and right-click the text for an existing entity, and then click the name of another entity.
For example, to make sure the car make in the text maps to the CarMake entity, highlight and right-click Ford, and then click #CarMake.
To use this training record to improve the artificial intelligence of the email bot, in the Review training data section, click Mark reviewed.