When the chatbot doesn’t have the answer: Tangowork Chatbot Accelerator v13

Dec 4, 2017

The Tangowork Chatbot Accelerator v13 improves scenarios where the chatbot doesn’t have the answer.

There are 2 main reasons why a chatbot might not be able to answer a user’s question: (1) it doesn’t understand; or (2) it understands, but doesn’t have an answer. In the Tangowork Chatbot Accelerator v13, we’ve added better support for the second situation with intent-specific “no answer” messages.

Imagine you have a list of bios for key staff members. A conversation might look like this:

User: Who is our director of learning?
Bot: Natasha Bowman is our Director of Learning


So far so good. But what if a user asks about a bio you haven’t added?

User: Who is our hr manager?
Bot: Sorry, I don’t have an answer for that.


In v13, we can do better! We can add an intent-specific no-answer message:

With this specific no-answer message, our chatbot’s response is much more useful:

User: Who is our hr manager?
Bot: I don’t have a bio for hr manager, but I have
all the bios for directors, VPs and the C-suite.

Similarly, we can specify a message to use if nothing matches the date or time specified by the user:

With the date-specific “no answer” message, we can have an exchange like this:

User: What’s on the schedule for december 8?
Bot: There’s nothing scheduled on december 8

This is a much better response than “Sorry, I don’t have an answer for that”!

Note that you can use {{value}} as an optional variable in your no-answer message. It outputs the value of the entity captured by the natural language processing.

To add variety to your no-answer messages, separate multiple messages with a pipe. One will be chosen at random.

Other changes

Quick access to sort order and list type. To directly access the sort order or list type of an intent, click it. It brings up a simple dialog box that lets you make a quick change.

More entity control with “Match exactly”. Sometimes typo-tolerance isn’t useful. For example, if the entity is a PO number or store number, we don’t want “212” to be treated the same as “121”. In cases like this, turn on the “Match exactly” switch.

More entity control with “Match entity only”. By default, all message content is searched for a match on the entity. Consider the user query “show me the bio for our ceo”. If you have a VP whose bio says “former ceo at XYZ corp”, it’ll match on “ceo”. However, if you turn on the switch “Match entity only”, it will only match bios that are specifically tagged with “ceo” as the job title.