Training the email bot to understand topics

For each topic added to the channel, you must train the email bot to interpret emails. By providing examples to the email bot and manually classifying them with a topic, the natural language processing (NLP) model can begin to understand these topics and classify similar emails. The email bot's classifications become more accurate and confident as it receives more training data. You can train the email bot in one of two locations: Prediction Studio, and the Email Bot channel. 

The training data tab within the Email bot channel enables the developer to collect, curate, and triage training data. When an email bot is first instantiated, the developer must add the training data to the channel and manually classify it. Once classified, the training data is marked as reviewed and added to a queue. The NLP model is then rebuilt to include the all-new reviewed records, as well as any existing training data that was already present within the model. Each time the model is rebuilt, a new F-Score is calculated that represents the model's overall accuracy across all topics.

1. In your email bot, click the Training data tab.
2. Click the More icon, and then select Add Records to add the training data. 
   Note: You can add example emails for the topics defined in the channel. A good starting point is typically 15-20 records per topic.

3. Click a record to see the classification details.
   Note: In the NLP analysis section, the Language model, Topic, and any entities present are displayed on the bottom tile. The training data text is displayed in the upper tile where any entities detected are highlighted.

4. Optional: Right-click the text in the preview pane to add or remove entities.
5. At the top, select the check box to select all of the records.
6. Click Mark reviewed to add the records to the build queue.
7. Click the menu, and then select Build Model to rebuild the NLP model. Once the model is built, the records are added and the model is updated.

If the model build is successful, a message that shows the new F-Score is displayed at the top.