One of the cool new features in Watson Conversation in IBM Bluemix is the ability to support a global audience by creating language specific versions of your bots. In this blog post I will give you some best-practices for creating bots in other languages.

To create a bot in a language other than English simply select the language from the pulldown for your workspace as shown below.

Select Watson language

Now that you have your workspace configured for your desired language you can start creating the intents and entities.

When you create your intents you need to avoid using message fragments. You should create complete sentences. For example, in my Spanish insurance bot I use the complete sentence “Localizar mi póliza” rather than the message fragment “Localizar mi” as shown below.

Watson intents

It is important to use complete sentences for Watson Conversation’s training engine to work correctly. You may be asking yourself at this point if you need to create a separate sentence for every synonym of “póliza”. The good news is that this is unnecessary. You simply list all the possible synonyms associated with “póliza” on the entities screen. So in my case I defined a few synonyms for “póliza” as shown below.

Watson entities

When you define your conversation nodes you simply indicate the intent and entity to use in the dialog. In my case I set my node to use the #policy-information intent combined with the @policy entity as shown below.

Watson dialog

I hope this post helps you quickly get going with creating multilingual conversational bots using Watson Conversation.

With the recent arrival of Watson Conversation on IBM Bluemix, developers are wanting to generate very sophisticated conversational bots that need to collect input and record state. In this blog post I will show you how to accept input in a conversational dialog and record state changes in your conversation.

After you have created your intents and entities in Watson Conversation you will want to start to create your dialog. The first question that developers immediately ask is how do you capture and then record the input of the user? To do this create a dialog node with a new condition called input.text as shown below.

Dialog with input.text condition

This node will record the user input and make it available as input.text. If you want to echo this input back to the user then you can use simply surround the condition with <? and ?>. So in this case we would use <? input.text ?> to repeat the input back to the user.

Now what do we do if we need to return the input value back to our calling application. How do we return the value? To return the input we will use a context object. To create a context object, switch to the “Advanced” view in the node by clicking on the three vertical dots in the right corner of the dialog node as shown below.

Selecting advanced node view

After you have the “Advanced” view displayed you will want to maximize the view by clicking on the “Maximize” button to make it easier for you to create and edit the JSON object as shown below.

Maximize node

Now add a section called context as show below in the editor and select a field name that you will use to refer to the input.text. In this particular case I created a field called policy and return the value that was input by the user. From your calling application you can simply access the policy field from the context object that is returned by the Watson Conversation API.

Advanced view

Now what happens if you need to refer back to your input.text later in your conversational dialog in one of your nodes. How do I access one of the fields from the context object? To do this let’s open up a maximized view of one of your nodes that needs to access one of your context fields as shown below.

Accessing context variables

As you can see, you can easily reference your context variable by preceding the context variable with a $. So in this case I access my variable through $policy. You will immediately notice that in the editor I have nothing in my context object and this is ok, as the context object I will be accessing is the one being used at run-time that was passed back to Watson Conversation through the API from your calling application. If I want to add more context variables I can do that at any time by simply adding more fields to the context object. Keep in mind that you can also replace existing fields by simply updating their values in the context object.

I hope this short post has helped you understand how to capture user input and carry state in your Watson Conversation bots.