1000 must have utterances for your chatbot using LUIS


My #chatbot is intended only to be pleasant, not to be philosophical or answer complicated political or social questions.  As companies create #chatbot to serve a purpose, they should also maintain a level of fun and make sure the #chatbot has at least a limited personality.  Utterances are questions which can be or have been asked to your #chatbot.  These questions or utterances are what get translated into an Intent and bound to an Entity.  Sure, one sentences or utterances can get close probability scores to multiple Intents and be bound to multiple Entities.  If that is the case, your code needs to be smart enough to handle them and respond in an intelligent, fun and witty way.  The code is where the magic and creativity will happen and once I get all the configuration setup here I will share some about the code.  #AI #ArtificialIntelligence #Cognitive #BotServices #Azure

If you have not read my other posts in regards to my #chatbot, please check them out here:

To upload the utterances, I did the following:

  • Check to see if any utterances already need my suggestion
  • Upload the utterances
  • Make the suggestions

Check to see if any utterances already need my suggestion

Login to your LUIS account here and click on the Edit button for the application you want to add the utterances to, as shown in Figure 1.

image

Figure 1, adding 1000 utterances to LUIS

Once on the main page for the application click on Suggest and make updates.  As seen in Figure 2, there is one and I need to select the Intent from the drop down and select any Entity which the word needs to be bound to.

image

Figure 2, updating and training LUIS Microsoft Cognitive Services

If there are no utterances then you would receive this message.

“There are no utterances from this application’s HTTP endpoint which haven’t been labeled already. Choose “new utterances” to provide your own examples, or publish your application’s HTTP endpoint to receive new utterances.”

Upload the utterances

Here is a text file with 1000 questions, ok, maybe only 994 or so.  Either way, go back to the main page where the list of your application are and click on Import Utterances, as shown in Figure 3.

image

Figure 3, import utterances into your #chatbot

Once uploaded I received this response window, Figure 4.

image

Figure 4, importing the #chatbot utterances

Click on the Edit button as shown previously in Figure 1 and click on the Suggest tabs as shown previously in Figure 2 as suggested in Figure 4.

Make the suggestions

After clicking on the Suggest tabs you will see a page similar to that shown in Figure 5.

image

Figure 4, maintaining the utterances in LUIS for a #chatbot using Azure Bot Services

And the last action is then to update them all by selecting the Intent from the drop-down and highlighting and selecting the bound Entity/Entities, similar to that shown in Figure 5.

image

Figure 5, #chatbot with LUIS, Microsoft Cognitive Services and Azure Bot Services

Once you go through and set the Intent and Entities for the ~1000 utterances, Train your application (see Figure 5 –> click the Train button on the lower left), Publish it again (see Figure 6) and go code it and make your #chatbot funny, bright, canny, clever, creative, intelligent, quick, resourceful or everything your would expect an AI personality with a SMART entity to be.

image

Figure 6, update and publish your trained AI LUIS model and change the world

Happy coding.  I have one more article I will write which will share some of the C# code I wrote to manage and respond to the questions.  Then I will leave the rest to you.  Rock!

Comments (3)

  1. Víctor says:

    Hi,

    Where is the text file link?

    Thanks,
    Víctor

  2. Dan says:

    Hi Benjamin,

    unfortunately, I can not find the text file containing the mentioned utterances in this article. Are they linked somewhere? Thanks for the great article!

    Dan

  3. cole says:

    Where is the text file link?

Skip to main content