Enrich Chatbot Entities With Patterns, Roles & Groups Using Rasa
…And Capture Concepts In User Input
Introduction
A conversational interface is an invitation you extend to the user to input their data in an unstructured and conversational manner.
It is imperative that the digital assistant is able to take the unstructured input from the user and ideally structure the input correctly on the first pass.
The less ideal situation is where the chatbot prompts for information already volunteered by the user.
Hence, forcing the user to input data in a structure acceptable by the chatbot. Keeping in mind, a structure not visible to the user.
The conversation to the left is a classic example of how, only the basic intent of the user is detected by the chatbot on the first pass.
The entities are disregarded and asked for, again; one at a time.
The way to approach this, is to handle it, as we do in human-to-human conversation.
City is an entity, but there are actually two types of cities, the from/departure city and the to/destination city. Hence the entity of city has two roles. And these two roles are arranged in a certain pattern including the date and possibly the mode of travel.
This is very close to how we as humans extract entities; detecting the patterns and identifying the roles.
The basic premise is to create roles for an entity, and then establish patterns in which those roles might be used.
These patterns then serves as a guide for the Conversational AI platform to detect patterns in user utterances. And subsequently useful conversational data.
Let’s have a look at how Roles and Entity Patterns can be implemented by using a simple banking example project.
So, different roles can be seen as subsets or types of the entity Cities.
Multiple roles can be added to an entity.
Contextual information is maintained for an entity in this way. We are also going to combine the roles defined for the entity with a pattern.
How To Implement Roles & Patterns With Rasa?
The best way to illustrate the use of entity roles and patterns in Rasa is by means of an example. Let’s look at a basic banking example.
In the nlu.md file, I have defined an intent to exchange money from one currency to another. This is an example of a user utterance with compound entities.
Within the user utterance, three entities are defined:
- Transaction
- Currency
- Date/Time
Here is an example of how to define entities with roles in the nlu.md file.
- intent: exchange
examples: |
- I want to [change](transaction) [US dollars]{"entity": "currency", "role": "change_from"} to [Euros]{"entity": "currency", "role": "change_to"} on [Friday](date_time)
- Can I [exchange](transaction) [euros]{"entity": "currency", "role": "change_from"} for [Rand]{"entity": "currency", "role": "change_to"} on [Saturday](date_time)
- give me a rate to [change](transaction) [pounds sterling]{"entity": "currency", "role": "change_from"} for [Japanese Yen]{"entity": "currency", "role": "change_to"} on [Monday](date_time)
- what are rates to [convert](transaction) [Australian Dollar]{"entity": "currency", "role": "change_from"} for [Russian Ruble]{"entity": "currency", "role": "change_to"} on [next week Thursday](date_time)
- I want to [change](transaction) [US dollars]{"entity": "currency", "role": "change_from"} to [Canadian dollar]{"entity": "currency", "role": "change_to"} on [Friday](date_time)
The entity of currency is broken up into two roles:
- Change From and
- Change To.
Here is a practical example of user input requesting to exchange money. This is performed using Postman and the Rasa NLU API.
The JSON output clearly define the recognized intent of exchange. With this, three entities are identified. The entity of currency has an additional value of role. In this case, the segmenting the currency changed from and to.
Group Entities
Entity groups allow for entities to be grouped together in different groupings. Using the banking example, I have created the entity tiers. There are 5 tiers; Silver, Gold and Bronze; grouped together. Constituting group one.
Platinum & Black are grouped together. Constituting group two.
This is an example of how to define entities with groups in the nlu.md file.
- intent: tiers
examples: |
- [Silver]{"entity": "tiers", "group": "1"} accounts have lower interest for saving than [Platinum]{"entity": "tiers", "group": "2"}
- [Gold]{"entity": "tiers", "group": "1"} accounts have lower interest for saving than [black]{"entity": "tiers", "group": "2"}- [Bronze]{"entity": "tiers", "group": "1"} accounts have lower interest for saving than [Platinum]{"entity": "tiers", "group": "2"}
And the example of a user utterance, mentioning two of the tiers.
The result is a JSON output with the entity tiers defined, and the grouping of the entity.
Conclusion
Instead of the chatbot forcing the user to structure input according an unseen pattern, the chatbot must create the structure from user input. And, have the capability to take the user input, identify intent, patterns, roles and intents.
The caution in using this feature from Rasa:
his feature is experimental. We introduce experimental features to get feedback from our community, so we encourage you to try it out! However, the functionality might be changed or removed in the future.
