Rasa Vs Dialogflow – Which one to choose?

There are many virtual assistant platforms out there, such as :

1. Rasa – Conversational AI-powered by open source
2. Dialogflow
3. Microsoft Bot Framework
4. Amazon Lex
5. BotKit

It’s obvious to get confused in such cases. If you are a developer/manager who wants to know what to take into consideration when comparing two chatbot frameworks, you have landed at the right place.

In this blog, we are going to compare two of the powerful chatbot frameworks i.e Rasa and Dialogflow and will make it easy for you to decide which one is best suitable for your project environment.

Dialogflow

Dialogflow is one of the most famous chatbot frameworks owned by Google. It was previously known as api.ai before it was acquired by Google in 2016.

Dialogflow is an NLU platform used by developers and non-developers to design and integrate conversational virtual assistance into mobile apps, web applications, as well as virtual assistance devices.

Rasa

Rasa is an open-source framework using Machine learning tools that can help chatbots to move one step forward. It has two main components – Rasa NLU and Rasa core.

Rasa NLU is particularly responsible for handling intents and entities whereas Rasa core is responsible for handling dialogues and context of the conversation.

Virtual Assistant types – FAQ assistant, Conversational assistant, Personalized assistant

When it comes to comparing two frameworks it totally depends on the type of bot you want to build.

FAQ assistance is the easiest type of assistant and can be built with almost all of the frameworks out there.

Talking about the conversational assistant, Rasa is best suitable as it can manage the context in the conversation in a better way than Dialogflow does. In Rasa dialogue and context is managed by stories. As most of the conversational assistance are contextual, context management is the backbone of the dialog management system. Both rasa and Dialogflow use slots to manage context. In Dialogflow, it’s a bit tricky to manage context but once you get a hold of it you can build a pretty good conversational assistant. But if there are complex conversations with more number of intents and entities it can make things difficult.

Personalized assistance is something which can improve itself and can handle context and conversation in a better way over time. Which can be possible by using machine learning-based Assistance.

SDKs and APIs availability

Rasa SDK provides tools we can use to write custom logic for our assistance.

Dialogflow has its own SDK available for different languages.

Let’s compare three components of Assistant development life cycle i.e. Development, Deployment and debugging.

Development

Learning Curve

For developers like you and me, it’s always important that the learning curve of any framework is short and we can start development as fast as possible. If there is no good documentation available for a particular framework then it doesn’t matter how good the framework is and how many useful features it provides.

Rasa’s documentation is easy to understand. It provides a getting-started guide for new users, which is absolutely a gem for newbies. It is easy to learn Rasa with available resources for beginners but not as easy as Dialogflow because of its complex architecture.

As Dialogflow is built, having both programmers and non-programmers in mind, its learning curve is quite short in comparison to Rasa’s.

Development Efforts

Development efforts and time generally depends on the project architecture, requirements, and expertise in the platform. The framework should be easy to use so that it can be built in a short period without compromising its quality.

As Rasa is an open-source platform it provides more flexibility to users to customize it. It requires a setup and understanding of a particular platform to get started.

Whereas Dialogflow provides an easy to use the framework and easy integration methods, which can help reduce development time as well as efforts when compared to Rasa.

Configuration Efforts

Rasa is a python library that’s why it requires installation as well as some basic set up to start assistance development, which can take up to 15-30 minutes based on the dependencies issues you have faced while installation.

Whereas Dialogflow doesn’t need any installation procedure. The only set up required is the webhook section which barely takes a couple of minutes.

In-built support

Having built-in features such as basic entities, small talks and demo assistance
supports developers to reduce time and efforts for developers. Dialogflow provides built-in support for that whereas Rasa doesn’t have built-in support such as.

Migration from other assistance platforms

Rasa supports easy migration from other chatbot platforms such as Dialogflow,
Wit.ai, and more.
Dialogflow provides migration to two platforms only i.e. Amazon Alexa and Google actions.

Deployment

When data security is concerned, most of the clients are concerned about their data and access to it. In the case of Rasa, data will be stored in the client’s own server as it’s on-prem software. So clients can have full control of their data.

Whereas it’s not the same in Dialogflow.. Dialogflow is a google cloud-based platform. The entire project and models will be stored in the cloud itself. We only have the ability to integrate backend custom logic in on-premise.

Debugging

After development, Debugging is the last but important part of any software development as it can eat up a lot of time. Till last year Rasa had the best debugging tools, but after the Dialogflow updated a new section called Validation – which is responsible for evaluating each intent and also provides feedback for any model which needs more training data or any kind of improvements.

Community support

Community support platforms are saviours to a developer. Dialogflow provides community support via its google group, Stackoverflow and Slack. While Rasa has its own forum and StackOverflow.

Conclusion

We hope after reading this blog you can now easily compare your project requirements with these two frameworks and choose the best suitable framework for your new assistance development project. Feel free to contact us in case of any queries.