Master Conversational AI with Rasa

Table of Contents

1. Introduction
2. What are Slots?
3. Defining Slots in Rasa
4. Setting Slots using NLU
5. Setting Slots using Custom Actions
6. Using Slots to Influence Conversation Flow
7. Configuring Slots for Conditional Conversation Flow
8. Including Slots in Training Stories
9. Slot Mappings in Rasa
10. Different Slot Mapping Types
11. Custom Slot Mappings
12. Types of Slots in Rasa
13. Text Slots
14. Boolean Slots
15. Categorical Slots
16. Float Slots
17. List Slots
18. Any Slots
19. Configuring Initial Values for Slots
20. Conclusion

Introduction

In this article, we will dive deep into the concept of slots in Rasa, one of the most powerful features that enable your assistant to Collect important pieces of information and use them in a specific Context. We will explore how to define and configure slots in Rasa, different ways to set slots using NLU and custom actions, and how to use slots to influence conversation flow. Additionally, we will discuss slot mappings, types of slots available in Rasa, and best practices for configuring and utilizing slots effectively.

What are Slots?

Slots are like a memory for your assistant. They allow your assistant to store and retrieve important pieces of information provided by the user during a conversation. For example, in a flight booking assistant, the user's destination would be a crucial Detail that can be extracted as a slot and later used by the assistant when making a booking.

Defining Slots in Rasa

To define slots in Rasa, You need to add them to the domain.yaml file of your assistant. Each slot requires the following information:

• Slot name
• Slot Type
• Parameter to determine if the slot should influence conversation flow
• Slot mappings, which define how the slot will be filled

Setting Slots using NLU

There are two main ways to set slots using NLU in Rasa. One way is to set slots Based on the values of extracted entities. This can be done by configuring slot mappings in the domain file. Another way is to set slots using specific values if a certain intent is predicted. The method for setting slots depends on how you configure your slot mappings.

Setting Slots using Custom Actions

Another way to set slots in Rasa is by using custom actions. Custom actions allow your assistant to connect to databases, make API calls, and extract information that can be set as slots. This method provides flexibility and allows you to set slots with dynamic values depending on the context of the conversation.

Using Slots to Influence Conversation Flow

Slots can be used to influence the flow of the conversation. You can configure slots to drive the conversation in a specific way based on the presence or value of a particular detail. For example, in a flight booking assistant, if the user provides the destination detail at the beginning of the conversation, the assistant can immediately check for available flights. On the other HAND, if the user doesn't provide the destination, the assistant will need to ask for it before proceeding. It's important to be mindful of when and how you configure specific slots to influence conversation flow and only do so if there is a valid reason for doing it.

Configuring Slots for Conditional Conversation Flow

To configure slots for conditional conversation flow, you need to include them in your training stories. Training stories teach your assistant how to drive the conversation based on the behavior of slots. By including steps where slots are set in your training stories and providing the details about the slots and their values at specific states of the conversation, you can teach your assistant how to handle different scenarios.

Slot Mappings in Rasa

Slot mappings are a powerful feature in Rasa that allow you to define how each slot will be filled. There are various types of slot mappings available, including from entity, from text, from intent, and from trigger intent. These mappings determine how the slots will be filled based on extracted entities, user messages, predicted intents, and form triggers. You can also Create custom slot mappings to fit your specific use case.

Different Slot Mapping Types

Rasa provides different types of slot mappings to cater to various scenarios. From entity slot mapping fills slots based on extracted entities. From text slot mapping uses the text of the latest user message to fill slots. From intent slot mapping fills slots based on predicted intents. From trigger intent slot mapping fills slots when a specific form is activated. You can choose the appropriate slot mapping type based on the behavior you want for each slot.

Custom Slot Mappings

If none of the predefined slot mappings fit your use case, you can create custom slot mappings using slot validation actions. These actions are a special case of custom actions and allow you to define how each slot should be filled. With custom slot mappings, you have more control and flexibility over slot filling and can handle complex scenarios.

Types of Slots in Rasa

Rasa offers several types of slots to store different types of information. Understanding the types of slots available will help you choose the most suitable slot type for the information you want your assistant to store and use. The types of slots in Rasa include text slots, boolean slots, categorical slots, float slots, list slots, and any slots. Each slot type has its own characteristics and usage scenarios.

Text Slots

Text slots are used to store any text values. They can be configured to influence conversation flow based on the presence or absence of a specific detail. For example, in a flight booking assistant, if the user provides the destination, the assistant can proceed with the booking process. If the user doesn't provide the destination, the assistant should ask for it. The value of the text slot itself doesn't influence the conversation, only the presence or absence of the detail.

Boolean Slots

Boolean slots are used to store information that can have two values: true or false. They can be configured to influence conversation flow based on the value of the slot. For example, if the user provides their user ID and it is authenticated, the assistant can proceed with specific actions. If the user is not authenticated, the assistant can respond differently and ask for additional details or inform the user about the authentication process.

Categorical Slots

Categorical slots are used to store values that can belong to a specific set of categories. They can also influence conversation flow based on the value of the slot. For example, in a restaurant booking assistant, the user might specify the price range they prefer, such as "medium". The assistant can use this detail to make the search for a restaurant more specific.

Float Slots

Float slots are used to store numerical values. They can also influence conversation flow. For example, in a restaurant booking assistant, the user might specify the radius within which they want to find a restaurant, such as "10 miles". The assistant can use this detail to optimize the search.

List Slots

List slots are used to store lists of values. They can influence conversation flow based on the presence or absence of a particular detail. The length of the list doesn't affect the conversation flow. List slots are useful when you want to store multiple values for a specific detail. For example, in a grocery shopping assistant, the user might provide a list of items to buy, such as "chocolate milk" and "cookies". The assistant can extract these details as entities under the same slot and use them to create a shopping list.

Any Slots

Any slots can be used to store arbitrary values that you want your assistant to have access to. These slots do not influence conversation flow. They are suitable for storing miscellaneous information that you want your assistant to remember and use in a specific context.

Configuring Initial Values for Slots

You can configure initial values for slots in Rasa. Initial values are specific values that are assigned to slots at the beginning of a conversation. These values remain assigned to the slots until they are reset by NLU or custom actions. Initial values can be used to set the starting point for the conversation or provide default values for slots.

Conclusion

Slots are a powerful feature in Rasa that enable your assistant to collect and utilize important pieces of information. By defining and configuring slots effectively, you can enhance the conversation flow and provide a more personalized and efficient user experience. Understanding the different types of slots, slot mappings, and best practices for utilizing slots will help you build intelligent conversational agents that can handle complex scenarios and provide accurate responses.

Highlights

• Slots in Rasa enable assistants to collect and utilize important information in conversation
• Slots can be set using NLU and custom actions, and can influence conversation flow
• Slot mappings allow you to define how slots will be filled based on entities, text, intent, and form triggers
• Different types of slots are available in Rasa, including text slots, boolean slots, categorical slots, float slots, list slots, and any slots
• Initial values can be assigned to slots to set starting points or provide default values
• Effective use of slots enhances conversation flow and improves user experience

FAQ

Q1: How can I set slots using NLU in Rasa?\ A1: Slots can be set using NLU by defining slot mappings in the domain file. Slot mappings allow you to specify how slots should be filled based on extracted entities or predicted intents.

Q2: Can slots influence the flow of the conversation in Rasa?\ A2: Yes, slots can influence the flow of the conversation in Rasa. By configuring slots to influence conversation flow, you can drive the conversation in a specific way based on the presence or value of a particular detail.

Q3: What types of slots are available in Rasa?\ A3: Rasa offers different types of slots, including text slots, boolean slots, categorical slots, float slots, list slots, and any slots. Each slot type has its own characteristics and usage scenarios.

Q4: How can I configure initial values for slots in Rasa?\ A4: Initial values for slots can be configured in Rasa by specifying a specific value for each slot in the domain file. These initial values are assigned to the slots at the beginning of a conversation.

Q5: Can I create custom slot mappings in Rasa?\ A5: Yes, you can create custom slot mappings in Rasa by using slot validation actions. Custom slot mappings allow you to define how each slot should be filled based on your specific requirements.