If Cloud was a term you were getting used to in the world of Information Technology, there is another jargon which has created a buzz lately. Cognitive Computing or Cognitive Services is the buzz that you hear everywhere. In this blog post I take a dive in to the world of Cognitive Computing using IBM Bluemix and in general the Watson Services on Bluemix. Let’s start the journey.

Cognitive Computing:

If you are not familiar with the term Cognitive Computing, don’t worry. We will figure it out together as we proceed. First thing I did was to check the dictionary of the web – Wikipedia. According to Wikipedia, Cognitive Computing is defined as below:

“New hardware/software that mimics the functioning of human brain and helps to improve human decision-making. Cognitive Computing is all about trying to build accurate models of how human brain/mind senses, reasons and responds to stimulus”.

Some of the real-world use cases where Cognitive Computing is applied are:

  • Conversations
  • Speech recognition
  • Sentiment analysis
  • Face detection
  • Recommendations

Cognitive Computing in IBM Bluemix:

On Bluemix (i.e. IBM’s Cloud), IBM is offering the Cognitive Computing services using what is popularly called as Watson APIs. Watson is nothing but a set of open APIs and SaaS products. Key features of Watson API & products are as follows:

  • Understanding all forms of data. This includes unstructured text, images, audio and video
  • Interact naturally with people. Conversation APIs help you build chat bots and engage in dialog
  • Learn & Reason. Use Machine Learning to improve subject matter expertise and offer recommendations

You get all the above features at scale. IBM makes sure that the Watson services scale as your business needs scale. IBM manages the scaling while you manage your business needs.

Watson APIs and Services:

Some of the Watson APIs and Services we will be considering in this post are listed below:

  • Conversation
  • Visual Recognition
  • Text to Speech
  • Language Translator
  • Tone Analyzer
  • Personality Insights

For a complete list of APIs & Service you can have a look at the Watsons product offering page here.


For the rest of the blog post I am assuming that you have a Bluemix account. If not, head over to https://console.bluemix.net/ – and create a free account for yourself. It will be easier to follow the rest of the post if you have your own Bluemix account and log in.

Conversation Service:

The first service we will look at is Watson Conversation Service. With the Conversation service, you can quickly build & deploy chat bots or virtual agents. These chat bots/agents can be made available across a variety of channels such as Mobile devices and messaging platforms. Let’s start by building a simple conversation bot using this service. Let’s get started.

Creating Conversation Service:

Log in to Bluemix console with your credential. Click on “Catalog” menu. Select “Watson” from the Services category. You will be presented with all the services available under Watson. You will see Conversation as one of the service offering. Click on Conversation.


Next, we will need to provide a name for our service. Click create once done.


Bluemix will go ahead and get the service ready in our console. To work with Conversation API, we will need to use a separate tool. Click on the Launch tool button.


Watson Conversation Tool:

In the Watson Conversation tool, we will need to first create a workspace for our app. This workspace is like a container for our app and will hold all the Watson artifacts for the app. Give a name for your container & a description and click Create.


Below is a screenshot of the Watson Conversation tool interface:


Creating Intents:

Next, we create Intents. An Intent is nothing but the purpose of a user’s input or actions that the user wants to perform. To keep it simple, I will create 2 intents – one for saying Hello and one for saying Goodbye. Click “Create new” in Intents tab.

Name the intent hello. To make the conversation service learn what is hello, enter phrases such as “hello”, “Hi” – anything that depicts the intent hello. Enter these phrases in user example.


Similarly, create two more intents named time & goodbye. And add user example which depict the intents time & goodbye.


Now we have 3 intents created – hello, time & good bye. Your intent tab should look like below:


Creating Dialogs:

Next, we will need to build a dialog. A dialog in Watson service is nothing but a conversation flow. We define the conversation flow in a logical tree format. Select Dialog tab and click Create to add new dialog.

We will have 2 nodes in the logical tree – “Welcome” & “Anything else”. “Welcome” node contains a greeting that is displayed when a user engages with the conversation bot. “Anything else” contains phrases which the bot uses when users input is not recognized.


Next, we will add 3 nodes to handle our hello, time and goodbye intent. Hello intent to greet the user, Time intent to handle user’s question on current time and Goodbye intent to handle exit.

Click More icon on the Welcome node and then select “Add node below”. Give a name for the intent. in this case we will name it Hello. Then select the intent #hello in the condition. Add a response phrase – let’s say “God day to you”. Click the big X to finish adding the node.


Similarly add nodes for time & goodbye intent. Provide a response phrase for the dialogs created for time & goodbye intent.


Try it out:

Now that we have finished adding the intent & dialogs – it’s time to test the intent recognition and dialog rendering. Click the chat icon found at the right hand top corner of the page. We will be presented with “Try it out” pane. Here we can enter inputs as if a user is typing it out. We will see that the Watson conversation service detects the right intent and the response phrase we have added for the intent is presented to the user.


With the above steps, we just finished creating a simple conversation with 2 intents – hello and goodbye. Now that our conversation service is in place, you can deploy this to any channels of your choice to create a bot.

Client Application to Consume Conversation API

Now that we have a simple conversation service set up, it’s time to create a client application which can consume it. I have chosen Node as the platform to write a simple client application. You will need to install node on your system to follow the code below. Let’s see what it takes to build the client app.

Create a new folder in your system and give it a name of your choice. Open a command prompt, navigate to your newly created folder and execute command – “npm”. NPM will ask you some question, provide answers and one done you will have package.json created. Next let’s install Watson-developer-cloud and prompt-sync package. Execute the following command:

npm install watson-developer-cloud prompt-sync –save

Watson-developer-cloud package will provide us the API required to communicate with our conversation service. Prompt sync package provides API to prompt users for input. Now, create a new file and name it index.js. We will add code in index.js to communicate with the service. Add the following code to initialize the Conversation API

var prompt = require(‘prompt-sync’)();

var ConversationV1 = require(‘watson-developer-cloud/conversation/v1′);

ConversationV1.URL = ‘<Watson Conversation Service URL>';

// Set up Conversation service wrapper.

var conversation = new ConversationV1({

username: ‘<your user name>’, // replace with username from service key

password: ‘<your password>’, // replace with password from service key

path:{workspace_id:'<your workspace id>’ }, // replace with workspace ID

version_date: ConversationV1.VERSION_DATE_2016_07_11


Watson Conversation URL, your username/password and workspace id can be found in your Watson Conversation Service tool. Click on Deploy icon and you will be presented with these details:


Next we start the conversation by sending empty message to our service. Add the following line of code:

conversation.message({}, processResponse);

We handle the response from the service in a function named processResponse. Let’s see the code for processResponse:

function processResponse(err, response) {
   var endConversation = false;
   if (err) {
     console.error(err); // something went wrong
   var message = “”;
   if (response.output.text.length != 0) {
         message = response.output.text[0];
   // If an intent was detected, log it out to the console.
   if (response.intents.length > 0) {
     console.log(‘Detected intent: #’ + response.intents[0].intent);
     if(response.intents[0].intent === ‘time’)
         message = new Date().toLocaleTimeString();
     if(response.intents[0].intent === ‘goodbye’)

  // Prompt for the next round of input.
         var newMessageFromUser = prompt(‘>> ‘);
             input: { text: newMessageFromUser }
             }, processResponse)

The code above is self-explanatory. We take user input and post it to our conversation service. Conversation service detects the intent from users input. We check the intent through the response object of our conversation service. If the intent was “hello” we show the response phrase from the service. If the intent was detected as time, we show the local time to the user. We prompt the user for input till we see that the intent entered by the user was good bye. Here is a screenshot of the dialog output with our conversation service:


Notice how the conversation service detects intent outside of the defined ones and can return an appropriate response.


Watson Cognitive services are a bunch of APIs which are available as services on IBM Bluemix cloud. We looked at one of the interesting service, namely “Conversation Service” in this blog post. This was my first time exploring the Cognitive services from IBM Bluemix and so far, the experience of using the service has been good. I could get up & running with the service without much hassle. As you have seen from the above sections, I was able to create the conversation service without writing a single line of code. Of course, the service I created was a hello world experience but if that’s made difficult – my impression on using the service would have varied to a large extent. Also, we saw that consuming the conversation service within your app is also super easy. With the node package, I could write 10 lines of code and create a simple chat experience in no time.

What I showcased was simple conversation but the potential of Watson conversation API is left to your imagination.

  • Kannan

    Nice Post