LinkedIn
Copied!

Table of Contents

Reviewing and adding training data for text predictions

Improve the performance of your text prediction by using the training data that customer service representatives (CSRs) add to a conversational channel as they work with incoming messages in the production environment. CSRs can give feedback on the outcomes (topics, sentiments, entities) that the text prediction detects in incoming messages, and correct any errors.

You can export the training data from a channel in the production environment, and then import it to the same channel in a non-production environment, for example, Development or Sandbox. The training data that you add to a channel becomes automatically available on the Training Data tab in the text prediction that is associated with the channel. You can then review and approve this training data, and rebuild the models in the text prediction with the new training data. Finally, the updated text prediction can be migrated and deployed in production.

  1. Export the training data from a channel in your production environment. For more information, see Transferring training data to another email bot and Transferring training data to another IVA.
  2. Import the training data to the channel in a non-production environment, for example, Development or Sandbox.
  1. Open the text prediction:

    1. In the navigation pane of App Studio, click Channels.

    2. In the Current channel interfaces section, click the icon that represents a channel for which you want to configure the text prediction.

    3. On the channel configuration page, click the Behavior tab, and then click Open text prediction.

  2. In the Prediction workspace, click the Training data tab.

    The training data that you imported to the channel is listed as pending training data.
  3. Optional:

    To add more training data, click Add, enter some sample text in the Text window, and then click Add.

    The text is added to the Pending training data list.
  4. Review the training data:

    You must review the training data before the system can add it to the respective model as feedback. All training data that customer service representatives add in the channel in App Studio or in the Pega Customer Service portal is visible on the Training data tab in the text prediction.
    1. In the Pending training data list, select a piece of training data.

    2. Optional:

      To edit the sample text, click Edit, modify the text, and then click Save.

    3. In the Topic field, enter the topic of the training data, for example, greeting.

    4. Optional:

      In the Sentiment field, enter the sentiment of the training data, for example, positive.

    5. Double-click and then right-click a word or phrase in the training data that represents an entity, and then select the corresponding entity type from the list.

    6. Select the Reviewed check box.

    Reviewing training data
  5. In the prediction workspace, click Save to save your changes.

Build your models to train them with the new training data. For more information, see Building models in text predictions.

    Have a question? Get answers now.

    Visit the Collaboration Center to ask questions, engage in discussions, share ideas, and help others.