In this blog series, we’ve looked at the pros and cons of Chatfuel and Dialogflow chatbot-building platforms as part of our development hack day. Now it’s time for the big finale, where Uchenna Okafor reveals what his team liked and loathed about using MS Bot Framework to create a chatbot – as well as which platform came out on top as the overall Redweb favourite.
What is MS Bot Framework?
MS Bot Framework is Microsoft’s comprehensive bot-building platform. It allows developers to create and manage intelligent bots that connect with users across a variety of platforms – whether that’s in an app, or on social media.
The extensive framework includes the Bot Builder SDK, Developer Portal, Bot Directory and Bot Connector. It also provides the Direct Line REST API, which allows users to host their bot in an app or website.
The Microsoft Bot Framework is commonly broken down into two separate components:
- Bot Connector
- Language Understanding Intelligence Service (LUIS)
The Bot Connector does exactly what it says on the tin, connecting bots with a vast selection of platforms. LUIS is a machine learning service that allows developers to build natural language into apps, bots, and Internet of Things (IoT) devices. It’s designed to enable developers to build smart applications that can understand human language and react accordingly to user requests. LUIS also offers a set of REST APIs to automate the application creation process.
The bonuses of MS Bot Framework
One of the main things we liked about the platform was how we could easily integrate with other Microsoft services and platforms. That meant we could make the most of LUIS, which would improve the quality of the chatbot by using natural language, while Azure Bot Service helped speed up development once we were up and running.
The framework also offers a certain degree of flexibility in how you build the bot. You can either use services like Azure Functions to power the data for your chatbot (which requires less code), or you could use your own data source (requiring more code).
As well as being open source, MS Bot Framework also enables users to share tasks with pair programming, which we felt were big bonuses when building with the platform. These features, alongside the extensive list of available integrations with platforms like Facebook, Skype, Slack, Telegram (and many more), makes MS Bot Framework quite the force to be reckoned with for those considering building a chatbot.
While a lot of MS Bot Framework was very comprehensive and impressive, not everything was perfect. This was not necessarily to do with the tools themselves, so much as the complexity in getting properly set up.
As a .NET framework, it requires a bit of knowledge to get up and running compared to the other chatbot platforms. Because of this, learning how to integrate MS Bot Framework with other Microsoft services to explore its full potential was quite complicated.
If you’re not familiar with Microsoft’s products, then you could find yourself struggling to get much value out of the framework without doing a lot of reading first. Of course we only had a day, so this limitation probably wasn’t helped by the time restrictions we placed on ourselves!
What did we learn?
MS Bot Framework is a really powerful platform – especially when paired with LUIS. We appreciated the freedom of being able to power the bot using existing Microsoft Azure products or by designing our own.
The integration of LUIS was also great because it allowed us to create a chatbot that could understand the context of the given text. For example, we were able to say, “Show me a list of cats” or “Show me cats” and it will understand that the intent was the same. This contrasts with other bot platforms, where users must be specific in their choice of words for it to understand the requests being made.
Overall outcomes of the hack day
The hack worked well and was a great opportunity to share our newfound knowledge with the rest of the agency. While the results and outcomes varied during the different teams’ demos, everyone enjoyed the day and came away with experience of a leading chatbot framework.
Comparing the frameworks themselves, we felt Chatfuel was probably best viewed as a prototyping platform. While results can be created quickly, the platform is a little unreliable and bots produced may be limited in what they can do. It’s also Facebook only, so isn’t especially scalable.
Both MS Bot Framework and Dialogflow are extremely powerful and capable platforms, but both require more work to get a production-ready working bot. When it came down to it, we had a clear favourite: Dialogflow was that much easier to use and explore, so we look forward to doing more with this platform in the future.
Here’s to more hacks in 2018!