Why still chatbots today? Is this a kind of revival of the good old days, when the little — often taciturn companions — popped up somewhere during chats or introductions to new programs? And are they the same people who, even after a patient conversation, only had one answer ready: “I'm sorry, I didn't understand your question.”?
No Chatbots are actually experiencing a kind of revival right now, but with new resolutions. Perhaps the time has even struck for those who were less smart and now eloquent companions. In any case, it's safe to say that the topic of machine learning was not as present over 10 years ago as it is today. And that is an important requirement for chatbots today.
Finally get rid of the forms
Many companies want a better user experience. Everything should be more intuitive to use and also save money and time. Be it in the area of customer service, shopping, ordering sushi or for more complex services, such as those found at banks and insurance companies.
So did one of our customers. His challenge: For his tailor-made software solutions, he must know the needs of his customers very precisely; after all, it is primarily individual software that our customer provides. Until now, the medium-sized company had collected forms for this purpose with questions about individual configurations. During the feedback process, some questions repeatedly turned out to be too complex to be discussed exhaustively in such a form.
And as you can imagine, it took some time for the forms to get back to our customer. The constant inquiring, processing and re-enquiring was an immense amount of time for him.
How nice if you could automate that, isn't it?
The Chatting Colleague
Chatbots have recently received a lot of attention from experts and companies. First of all, this is due to changing technological possibilities. They make it possible to evaluate larger amounts of data and to react not only to texts, but also to speech or images.
Second, it is due to the ever-improving models for Natural Language Understanding (NLU). They make it easier to interpret human language. This is the brain of chatbots, so to speak. We wanted to try the experiment and our customer jumped on the boat. The chatbot should be able to solve the problems mentioned earlier: save resources and reduce complex processes to a minimum of human work.
“It makes a difference whether they are 95% or 99% accurate. ”
DAVID KAISER, Amazon
about chatbots
The rhetoric test: Comparing Rasa vs. Microsoft
A rough analysis based on the customer's specified requirements revealed that there are two chatbots on the market that are suitable. That was Rasa and Microsoft (MS) Framework Bot.
We have taken a closer look at the two of them.
commonalities
- First of all, the positive thing about both is that when it comes to hosting, you have the freedom to choose between cloud or on premise
- Both frameworks are currently of interest to many companies and have a broad community
- Both frameworks are GDPR-compliant With LUIS.ai and Rasa NLU, both providers provide powerful NLU tools
- Numerous framework samples in GitHub are available for all documentation is available
- There are interfaces for numerous external services, such as messaging tools, collaboration applications or websites
differentials
prize
Rasa: The bot is available as an open source version, which can also be used commercially thanks to an Apache 2.0 license. However, the full version is chargeable.
MS Bot framework: Based on the pay per use principle, but easy to order/unsubscribe in Azure
programming language (s)
Rasa: python
MS Bot framework: Follow .NET or JavaScript, Python and Java
Application width
Rasa: text interpretation
MS Bot framework: Text, voice, and image interpretation
complexity
Rasa: It is easy to use and has a great user interface, plus video tutorials and the easy-to-learn programming language python The bot is user-friendly
MS Bot framework: The framework is easy to implement, but it's harder to get your bearings. This is due to the less intuitive user interface, which takes some getting used to
limitations
Rasa: Preferably own hosting, hosting in the cloud is a bit more cumbersome.
MS Bot framework: A major drawback is the big lock-in effect. A separate implementation without Azure is also very complex. But the connection to Microsoft apps, such as Teams or Skype, works smoothly.
For us, Microsoft was ahead
It must also be said that, as a gold partner, we are already at an expert level in the Microsoft universe and we therefore understand the technological conditions more quickly. Our customer also had a lot of experience with Microsoft. But that wouldn't have prevented us from implementing Rasa for our customer as well. At the end of the day, the C# programming language, which both we and our customer are good at, tipping the scales. The handover and subsequent maintenance have made it easier for us and the customer. And as a bonus, we were impressed by how quickly you can create a chatbot using an Azure account.
In addition, Microsoft offered more functionalities that could be retrofitted later if required. The potential to upgrade graphical or language models later speaks for the flexibility of the Microsoft solution. And in the end, these aspects have also outweighed the sometimes “usual Microsoft” cumbersome usability.
