Skip to content

Latest commit

 

History

History
78 lines (55 loc) · 3.28 KB

09-entity-resolvers.md

File metadata and controls

78 lines (55 loc) · 3.28 KB
Raw

Entity Resolvers

This tutorial shows how to use Entity Resolvers in Conversation Learner

Video

Entity Resolvers Tutorial Preview

Requirements

This tutorial requires that the general tutorial Bot is running

npm run tutorial-general

Details

  • Resolver Type is an optional property of Custom Entities.
  • Entity Resolvers utilize the power of the pre-trained entity recognizers in LUIS to provide additional detail and clarity for your Custom Entity.

Steps

Start on the home page in the Web UI.

Create the Model

  1. Select New Model.
  2. Enter Entity Resolvers for Name.
  3. Select Create.

Create a pair of Entities

  1. Select Entities in the left panel, then New Entity.
  2. Enter departure for Entity Name.
  3. Select datetimeV2 for Resolver Type.
  4. Select Create. Dismiss the informational popup by selecting OK.
  5. Repeat steps 1-4 to create a second Entity named return with datetimeV2 resolver type.

Create a pair of Actions

  1. Select Actions in the left panel, then New Action.
  2. Enter You are leaving on $departure and returning on $return for Bot's Response....
    • IMPORTANT - When typing in $[entityName] you need to hit enter or click on the entity in the drop down otherwise Conversation Learner will consider this to be text instead of an Entity.
    • Notice that the Required Entities field will also get these Entities and they cannot be removed. This prevents this action from becoming available until both required Entities are present.
  3. Select Create.
  4. Select New Action to create a second Action.
  5. Enter When are you planning to travel? for Bot's Response....
  6. Enter departure and return for Disqualifying Entities. These tell our Bot to NOT take this action if either of these Entities contain a value.
  7. Select Create.

Training

  1. Watch the Training: [Status] in the upper left corner for Completed.

    • You can click the "Refresh" link if this takes too long.
    • Training status "Completed" is necessary so that our Entity Resolvers work when we train the Model.

  2. On the left panel, click "Train Dialogs", then click the "New Train Dialog" button.

  3. Type in the first user utterance, "book me a flight".

  4. Click the "Score Actions" button.

  5. Select the response, "When are you planning to travel?".

  6. As the user, respond with, "leaving tomorrow and returning Sunday next week".

    • Notice how Conversation Learner has detected two "Pre-Trained date" Entities in that user turn.

  7. In the "Entity Detection" panel, select the text "tomorrow" and label it as "departure"

  8. Also label the text "Sunday next week" as "return"

  9. Click the "Score Actions" button.

    • Notice how the "Memory" pane contains your departure and return dates.
    • Hover over each one and observe how the Entities are date objects which clearly capture the actual calendar date as opposed to "Sunday" or "tomorrow".

  10. Select the "You are leaving on..." Bot response.

  11. Click the "Save" button.

Next steps

[!div class="nextstepaction"] Enum entities