Pattern for gathering information in IBM Watson Conversations

In a previous post, I shared some thoughts and an approach to Recursive questions with IBM Watson’s Conversation service. The post detailed how we can collect a piece of information from users in a way that allows for that information to be provided in the initial utterance or as a response to a question from our bots.

In this post, I will extend this example and demonstrate a pattern to collect multiple pieces of information while maintaining flexibility and eliminating duplication. You can download the workspace source code from my Github account.

What are we trying to do?

The scenario is that we are creating a bot for a cloud service provider which will enable users to open a new account. Our Acme SaaS company provides three account types (Bronze, Silver and Gold) and two subscription types (Monthly and Annual). We have two entities for Account_Type and Subscription_Type and an intent for Open_Account.

I find it helpful when designing processes to be serviced by bots to map these out logically, I like to do this as follows:

Intent Entities Resolution
Open_Account Account_Type
Subscription Type
Account Opened

Creating the above mapping, particularly for complex use cases will help you identify where intents can be merged. If you get two intents with different Entity conditions but the same resolution you should merge the intents and disambiguate on the Entity values.

The above mapping identifies that we need to know the Account Type and Subscription Type before we can fulfil the request and open an account. The following diagram shows the strategy we are going to implement to fulfil our open account request.

The important thing to note from above is that we can describe the collection of the information in a recursive loop. Within the loop if a required entity is provided we save it to context, if additional entities are required we prompt for those and repeat. Once all requirements are satisfied we can fulfil the users request.

How do we do that?

To achieve this, with the dialog features of the Watson Conversation service, we will use a combination of @Entity and $Context conditions and the “continue_from” features in a single level within the dialog. Let’s take a look at the first few nodes.

screen-shot-2016-11-18-at-14-17-40Our first node matches on the intent for Open_Account and uses a “continue_ from” to go immediately to the check for @Account_Type. This node will fire if the user’s input includes a valid @Account_Type and uses it a “continue_from” to go immediately to the next entity check.

Looking at the advanced tab of @Account_Type or $account_type node, we can see that it returns no output and sets the account_type variable on the context object. We will use that context variable later to fulfil the request but also to make sure we don’t ask the user for the info twice.

{
  "output": {},
  "context": {
    "account_type": "<? @Account_Type ?>"
  }
}

Moving down the node stack at this level we do the same for Subscription_Type which if it’s present we will again use the “continue_from” and move on to check each context variable.

screen-shot-2016-11-18-at-17-55-05Our next few nodes have conditions that check if a context variable doesn’t exist and if not, prompts for the user to provide the entity. Once we have checked for $account_type and $subscription_type our final node fulfils the request (or at least says that it will).

Context variable checks and fulfilment node

Notice the use of <? $account_type ?> in the output of the last node, this inserts the values of the information collected into the response to the user.

Before wrapping up I’d like to point out one more thing you can’t see in the image above, but the nodes with conditions on the context variables (e.g. !$account_type) have complex output that uses a feature called sequential selection policy. Let’s take a look at the advanced json first then I’ll explain what it does.

{
  "output": {
    "text": {
      "values": [
        "I can help you open an account...Which option would you like?",
        "Sorry I didn't understand that...Gold account?"
      ],
      "selection_policy": "sequential"
    }
  }
}

When this node fires (i.e. when the $account_type variable is not set) it will output the first item in the values array “I can help you open an account…”. If the user fails to enter a valid @Account_Type then it will output the second item in the values array “Sorry I didn’t understand…”. With this approach you can re-prompt the user without having to introduce new nodes.

What does it all mean?

To conclude I want to point out the ways that we can interact with the Open_Account flow. If I were to ask “I’d like to open an account”, the bot would ask “what type of account?”.  If I then say “bronze” the bot will ask “monthly or annual payments?”. If I then say “monthly” the bot will fitful the request. However I could also say “I’d like to open a bronze account” in which case the bot will just ask “monthly or annual payments?”. If I say “I’d like to open a Bronze account with monthly billing” the bot will simply fitful the request.

Using this pattern we can allow the user to provide all, some or none of the information we require in the initial request. With a few nodes and one level of evaluation we can collect the information we need and even re-prompt the user where required.

As a reminder you can download the workspace I used to test this pattern on my Github account. Just download it and use the import feature in the conversation tooling to see a working example for yourself.

Recursive questions with IBM Watson Conversations

IBM Watson released a new API earlier this year called Conversations which enables developers to build cognitive conversational solutions like chatbots and virtual assistants.

A common scenario when creating a chatbot is the need for clarification questions, for example you are creating a bot that helps users open a bank account. The user might say “I’d like to open a current account”, the bot understands that the user intends to “open account” and the account type is “current account” (AKA checking account in the US). What if the user says “I’d like to open an account”? In this case, in order to help the user, we’d need to know the type of account they want to open.

There are a number of way to approach the collection of the required information in a chatbot. The image below shows an approach of splitting the conversation based on the information provided.

Screen Shot 2016-08-30 at 13.07.41

If the users intent is classified as “Open_Account” and an entity of the type “Account_Type” is found then the first node will fire. If however the intent is classified as “Open_Account” but there is no entity then the second node will fire. This approach will work just fine for these examples but if we have many more pieces of information we require to complete a request, then this approach will be more difficult to maintain.

One of our goals when designing dialogs should be to reduce duplication and provide flexibility to the end-user. Reducing duplication will lower the cost of maintaining solutions, while flexibility in the way users can interact will provide a much better user experience. The following image shows a better way to approach this flow.

Screen Shot 2016-08-30 at 14.31.30

The key to this approach lies in utilising the “continue from” features of the Conversations service so the flow will respond intelligently and intuitively to users.

The first thing to note is the “Continue from…” node at the top of the second column of nodes. This tells the dialog to continue straight from the initial node with the #Open_Account condition to the next level. Note also that the output of the first node is empty, this is required as we don’t want our bot to say anything we just want it to move on and evaluate the second level of conditions. You can check that the node will not output anything by making sure that the output is set to {} in the advanced view as shown below.

Screen Shot 2016-08-30 at 14.07.00

The second thing to note is the second column of nodes. The first node has a condition on “Account_Type”, the second has a dummy condition of “true” (This always evaluates to true). This tells our dialog that if the user input has an account type then go ahead and fire the first node, if not ask the user what type of account they want.

The last thing we add is a new “Continue from”, the last node which is set to “User input”. This tells the dialog that every time we hit this node return the user input and evaluate this stack again (i.e Recursive).

This approach ensures our bot gets the information it needs to continue through the conversation but it also locks the user into the flow. In a future post I will explain techniques for allowing users to break out of the flow which is important when our bot doesn’t understand or when the user wants to talk about something else.

Accessing the synonym mentioned with IBM Watson Conversations

A couple of my clients have asked me how to access the synonym mentioned by a user when working with the IBM Watson Conversation service. Accessing the matching entity is well documented but it wasn’t apparent how to get the actual word or words mentioned by the user that the entity was matched on.

Having asked the question internally and received the answer I thought I would share in lieu of the documentation being updated. The answer is actually in the documentation here but without explanation so my clients and I missed it.

The scenario is pretty common and straight forward, lets say you have an entity called Account_Type with an entry for Bronze with a synonym of Basic. Your dialog flow for opening an account asks the user what type of account they would like to open, to which they say “I’d like to open a basic account please”. At this point your dialog is configured with:

“Okay, I’ll open a @Account_Type for you now.”

1*l9WzyyIDJpvhJdLbBPTFGA
Example of outputting a matched Entity

The result will be a replacement of the @Account_Type placeholder with “Bronze”. This can be confusing for users and communicates a sense of inflexibility that is less than ideal. Instead we can access the synonym mentioned by the user with:

<?entities["Account_Type"][0].literal ?>

We configure our output as:

Okay, I’ll open a <?entities["Account_Type"][0].literal ?> for you now.

1*Epu1qC9hx3i_G4rAoUCDgA
Example of outputting the synonym used

The result will now be that the placeholder is replaced with the word “basic”.

Writing what I think, the way I think it

IMG_0333

I’ve been meaning to get started blogging for some time now and have finally decided that I simply need to get going. I’ve read the “How to start a Blog” guides several times, feeling inspired and confident after each iteration but something keeps holding me back.

Having thought about what’s been stopping me, I decided it’s something to do with a belief that what I write should be written perfectly. I’m no author, I’m far more comfortable speaking about what I think than writing it down.

A few weeks ago I was talking to a stranger on a plane, travelling back from Berlin, I explained to this stranger that I was in Berlin to give a talk. After a minute or two discussing the nature of my talk and my job they asked: “So I guess you blog a lot?”, “No, I’d like to but haven’t yet” I answered. I explained my reluctance to dive in and get started to which they responded “As someone who blogs quite a lot I’d give you this advice, just do it, don’t think about it, just write. Write like you talk, it can be your thing”.

So here I am, just writting what I think, the way I would say it to a stranger sat next to me on a plane, I hope that’s okay with you.