AI for the Enterprise

Register for a Watson webinar on how chatbots can tackle curveball questions

Share this post:

Key Points
– Chatbots are designed to understand and communicate with users — however, questions often cover subject matter that the chatbot hasn’t yet learned
– Find out how your bot can get the answers to questions that it hasn’t been trained on by combining Watson Discovery Service and Watson Conversation Service
– Watch our free webinar on “Leveraging the long tail to build chatbots for a more engaging user experience.”

Watch the replay video of our webinar


Please join us on Wednesday, July 12 at 1pm ET
Speakers: Joe Scherping, Associate Offering Manager – Watson Platform and Chris Desmarais, Offering Manager, Watson Conversation
Watch the on-demand webinar today

In this technical webinar you’ll learn how to build a chatbot using Watson Conversation and extend its function using Watson Discovery’s cognitive search to answer questions beyond modeled intents. The app lets users ask about Conversation’s technical documentation and returns answers and relevant technical documents using the language and context of questions to retrieve the most appropriate answers for the more complex inquiries. Find out how to use Watson Conversation with Watson Discovery together to build chatbots that create more engaging user experiences.

Extend your chatbot capabilities with an insights engine

Given the abundance of accessible information and channel options, people are looking to have their complex and less commonly asked questions answered as quickly as possible. Whether they are looking to change their billing information on a customer service call or for great restaurant recommendations in a new city, customers want personalized responses right at their fingertips. Chatbots can be trained to respond with useful solutions to these questions, however, at times inquiries may go beyond a chatbot’s modeled intents.

When you think of chatbots, you can categorize responses broadly into two groups that we refer to as common short-tail responses and more unique long-tail responses. Short-tail responses recognize simple questions and commands such as, “Can you turn on my wipers?” Conversation is trained to identify specific intents and provide solutions to these short-tail questions.

There are also long-tail questions, such as “How do I check my tire pressure?” that are outside what the chatbot is trained on and therefore don’t have defined intents. In these instances, the chatbot must refer to other content sources, such as FAQs or a manual, to successfully answer the user.


Learn more about combining Conversation and Discovery from the experts. Watch our On-Demand “Build with Watson” webinar.

Try Watson’s APIs to build your next smart app or solution in minutes

Associate Offering Manager - Watson Platform

More stories
May 24, 2018

New Watson Assistant updates to provide GDPR support

IBM is committed to providing our clients and partners with data privacy, security and governance solutions to assist you on your journey to become GDPR ready. To that end, learn how Watson Assistant has introduced a few new capabilities to help our clients along their GDPR compliance journey.

Continue reading

May 18, 2018

With Watson – an AI community to learn, grow and transform business

With Watson is a customer success program that brings a community of people, from individuals to enterprises, interested in learning and/or building with AI together. It provides exclusive access to AI marketing and technical experts to help accelerate your AI vision while getting the most out of Watson and the IBM network.

Continue reading

April 30, 2018

How KPMG uses AI to empower their auditors

KPMG looks to using Watson AI to drive higher-confidence outcomes, greater insight, and employee empowerment, and we are committed to investing in resources that enable a high-quality audit.

Continue reading