Learn how to build Watson Conversation service apps that can interpret the tone of the user. Also, discover how to integrate Watson services with back-end systems.

In this video:

IBM Watson Senior Software Engineer Dan O’Connor takes you on a short introduction to changes in the customer care industry (from human CSRs to automated, augmented ones) and to the Watson Conversation and Tone Analyzer services before launching into the demonstration on how to use them to set up a chatbot that can recognize the tone of the user.

Dan notes you can set up a bot that only communicates on a single social media channel or one that communicates across several channels.

The average human agent would easily understand if a user called up and said:

“I am frustrated. I haven’t been able to login to your online billing system.”

The virtual agent must break down the request into component parts to be able to understand it. In this case, component parts would be intents, entities, and tones.

The Conversation service breaks down the exchange into intents; in other words, what it will take to resolve the customer’s issue. The next key piece of information the service needs to extract is the entity the user provides; the “noun” of the situation. Also at this point, the channel the customer is using is also providing context to the agent: For example, if the user’s records show they always access the entity via a smartphone, the system can make all of its responses based on the context of a mobile user.

Finally, and this is where Tone Analyzer goes to work, the agent (through back and forth dialogs) assesses the tone of the user conversation and hopefully, responds with the appropriate emotional tone.

Dan explains the models that Tone Analyzer uses to categorize a customer. It employs the Big Five model – of user traits like agreeableness, conscientiousness, extraversion, emotional range, and openess – coupled with a 12-attribute Needs model (of 12 characteristic user needs like excitability, curiousity, etc.) and the Values model which characterizes a user based on a set of values like self-transcendance or conservativeness.

Resources for you

Building with Watson

The one-stop shop to jumpstart your cognitive application development.

Cognitive Computing

Create apps that accelerate, enhance, and scale the human experience.


A transformative way to let users interact with your organization.

Join The Discussion

Your email address will not be published. Required fields are marked *