Voiceflow offers you the ability to import and export the native data direct from/to your production NLU Platforms. The benefits of utilizing your actual NLU model data in your Voiceflow Assistants are:
- Kick-start your designs by leveraging your existing NLU data building blocks for your Dialog Designs
- Maintain parity between the NLU data in your designs and prototypes, and what is being used in your production Assistant
- Speed up the hand-off of your designs to development by providing production-ready data
Types of NLU Integrations
NLU Integrations come in two forms on Voiceflow - manual exports, or automatic uploads. The types we offer will vary per NLU depending on which types of data transfer they offer.
1-Click Uploads
With these we offer a direct API integration between Voiceflow and the NLU Platform that will allow you to connect your account, and upload your NLU model with the click of a button.
Currently, our 1-click uploads only work in a single direction - sending data from Voiceflow to your NLU Platform.
Manual Export+Imports
For these Integrations we will create export files pre-formatted for upload directly into your NLU platform, and consume the export files that your NLU platform generates.
Integrations Supported per NLU
We currently support the following integration methods per NLU platform:
Platform | Upload | Import | Export | File Type |
Amazon Alexa | ✅ | ❌ | ✅ | JSON |
Amazon Lex V1 | ❌ | ✅ | ✅ | JSON, ZIP |
Dialogflow CX | ✅ | ❌ | ❌ | JSON, ZIP |
Dialogflow ES | ✅ | ✅ | ✅ | - |
IBM Watson | ❌ | ✅ | ✅ | JSON |
Microsoft LUIS | ❌ | ✅ | ✅ | JSON, ZIP |
Nuance Mix | ❌ | ✅ | ✅ | XML |
Rasa | ❌ | ✅ | ✅ | YML, ZIP |
Salesforce Einstein | ❌ | ✅ | ✅ | CSV |
Voiceflow NLU | - | ✅ | ✅ | CSV |
What data is included in a Voiceflow Assistant
For these integrations, it’s important to know what data gets created on our platform as part of an assistant:
The three main data types included in a Voiceflow assistant are:
- NLU Data - this is all the data needed to train an NLU. This includes intents, utterances/training phrases, entities and entity values
- Logic Data - this includes external data sources, variables used to handle data, and conditional logic that defines different paths in the dialog design
- Response Data - this includes all content of a response (text, speech, visual, etc.) that an assistant would provide to a user
What data is included in an NLU Integration
Our Integrations are focussed on passing the NLU Data only to your NLU Platform. While some NLU Platforms will house your Logic Data and Response Data, we will not pass that to your NLU platform as part of our exports or upload, or consume it on an import.
The specific data types contained within your NLU Exports and Uploads will vary per platform, so here is an overview:
Platform | Custom Intents | Built-in Intents | Utterances | Custom Entities |
Built-in Entities |
Entity Synonyms | Fulfillment Webhook |
Amazon Alexa | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ |
Amazon Lex V1 | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ |
Dialogflow CX | ✅ | ❌ | ✅ | ✅ | ❌ | ✅ | ❌ |
Dialogflow ES | ✅ | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ |
IBM Watson | ✅ | ❌ | ✅ | ✅ | ❌ | ✅ | ❌ |
Microsoft LUIS | ✅ | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ |
Nuance Mix | ✅ | ❌ | ✅ | ✅ | ❌ | ✅ | ❌ |
Rasa | ✅ | ❌ | ✅ | ✅ | ❌ | ✅ | ✅ |
Salesforce Einstein | ✅ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ |
What is a Fulfillment Webhook
Some NLU platforms allow you to set a fulfillment webhook for your Assistants that will call a 3rd party endpoint operate your Dialog Management, including:
- Managing state for your conversation
- Determining the logic paths the conversation will follow
- Delivering the appropriate response back to the user
This function can be handled by Voiceflow's Dialog Manager, which enables you to easily utilize the other Logic Data and Response Data from your assistant directly in your production Assistant:
In the instances that can handle a fulfillment endpoint, we will set it. If you do not want to utilize this, can you disable.
Here is a quick introduction to the Voiceflow Dialog Manager:
Comments
0 comments
Please sign in to leave a comment.