Watson Conversation API
You may know, Watson has a Watson Conversation API. This api can be used to create chatbots, and manage customer complaints. My favorite way to use conversation api is using it as a complaint management tool. (I’ll be writing a tutorial about it, we will just trigger our complaint management tool’s endpoints)
What is Watson Conversation API
IBM’s Watson Conversation API is part of IBM’s Watson platform, that enables you to create conversation between man and machine a.k.a between your customer and customer relationship goals.
As i mentioned before, i’ll help you to optimize your processes, teach you new way to increase customer satisfaction, customer happiness and your income. I am going to transform your business from scratch to cognitive.
Use Watson Conversation API to Help Customers
Now we are going to create a conversation to help customer complaints. I have to note that before getting started, this tutorial based on this tutorial. Now log in to Bluemix Console. Navigate to Catalog, click on Conversation, fill the form and create a new conversation instance.
Click on Launch Tool. Create a new Workspace, it’s name is not important, just create a new workspace. After Watson creates new instance click on get started. Now you see 3 link on top menu.
Let’s explain them first.
Intent: Intent is a key for Watson to understand what does input mean. Watson processes intent.
Entity: An input may be about something, for example you sell different kind of electronics, you can define the goods and help your customer based on your product.
Dialog: Dialog is a kind of flow chart, that describes the way you handle customer complaints and questions.
For a second, think about a customer dialog, it starts with greeting, then robot ask about problem, then take actions to solve customer’s problem or save the praises, and say goodbye to customer. To create this scenario we are going to define #greeting intent first.
Now i think a little bit explanation required. Watson is a kind of computer that learns like a human. Think about how you learn maths, first you learn about a rule. Then you make lot’s of practices to improve your knowledge and skills on the topic. Watson learns just like you. More examples makes Watson gives better results therefore more examples means better results.
As you can see here, we are starting with giving keywords for Watson, then Watson, learns those keywords and creates an understanding about what is a greeting, when greeting intent fires.
Now continue with creating a complaints intent.
As you can see, a few complaints defined.
Now, create a new intent, name it byes, to define how may a customer say goodbye.
Our intents are defined now! We can create a dialog flow. navigate to Diaglog tab and click “Create“.
I simply created a flow to handle a customer. You see, our intents are defined on the dialog. Anything else is the path that Watson respond for input’s that didn’t understand clearly.
Here we come to the best part, we are going to test our robot. There is a chat icon, on top right of the screen. Click on it.
I started to test with random strings. Watson told me that, it didn’t understand the input. Then i said hi to Watson.
Watson identified it is a greeting, and responded well.
I told about a problem with my product.
OMG! Watson understand it although we didn’t define anything about product and glass.
Then i said bye to Watson.
Improving Your Robot!
Now we are going to make a little addition. First of all go to entities and create two entities.
Then go back to dialog and create a continue from condition.
And mark your entities after your condition. After this changes click test button. Probably, you will see this:
That means Watson Conversation API is training on your entities and conditions. In addition, you don’t have to do anything to train, Watson Trains automatically.
When Watson completes training, you see this message. And now we are ready to go!
Due to, conversation API is stateless we have to test our scenarios one by one. Start with car scenario.
Then clear and test your phone scenario then:
You created your first customer complaint bot under ten minutes. Hence, you can create your own robot to make your customers happier!
I’ll write about using Watson’s conversation API with node.js and socket.io later.