Predict customer needs and behaviors by using text predictions in your conversational channels (8.6)
Route conversations and populate case properties by running a text analysis on the messages in your conversation channels, such as email or chatbot. For example, you can route incoming emails to the appropriate department based on the topic of the email.
You can now configure text analytics for your channels through text predictions. This new type of prediction is available in Prediction Studio and replaces the text analyzer rule in Dev Studio. A text prediction is a powerful tool for defining and monitoring the outcomes that you want to predict by analyzing the text in your channels. The system automatically creates a text prediction for every channel that you configure in App Studio.
Text predictions are preconfigured to suit the channel type. For email channels, the associated text predictions include the default email parser. For chat channels, the text predictions are preconfigured with the small talk model and a set of entities.
Text predictions provide the following benefits:
- Consolidate the AI for conversational channels so that you can configure text analytics for your emails and chats in one place.
- Reduce the time to go live for your email bots and chatbots.
- Support intuitive monitoring of outcomes to uncover issues and improve performance.
- Facilitate the training of models to detect topics, sentiment, and entities in text that comes through the channels.
- Support an efficient asynchronous process that builds the selected models in the prediction.
- Offer a convenient way to test predictions.
You can open the associated text prediction from the Behavior tab of a channel in App Studio. In the text prediction, you can manage the text analytics for the channel by performing the following tasks:
- Select the languages that you want to analyze.
- Define the outcomes to predict:
- Configure topics to route your conversations.
- Configure or select existing entities, such as names and numbers, to autopopulate case properties.
- Train models to drive your business outcomes.
- Build and test your models before you deploy the updated text prediction to your production environment.
- Monitor performance of your outcomes for the selected languages and dates.
See the following video for step-by-step instructions for creating a chatbot channel, accessing the text prediction, and configuring entities with machine learning and keywords to detect flight numbers and airport names in the incoming chat messages:
For more information, see Analyzing messages with text predictions.
Previous topic Manage customer records in Customer Profile Designer (8.6) Next topic Replace models in predictions and migrate changes to production with MLOps (8.6)