Machine learning (unsupervised) and supervised¶
This feature is useful in case your assistant uses natural language.
To help the machine understand the demands of your users in your domain, it needs practice phrases. Let's draw a parallel with image recognition, to identify a cat, we need variants of photos of cats marked "Cat" ... In natural language, to understand a "I would like to pay my bill of 20 €", it You need variants of "I would like to pay my 20 € bill" ... marked with 'bill.pay' for example. This is called supervised learning. Your assistant will not invent anything, he will try to rank a request among what you have taught him with a confidence score.
You have two choices:
- Or link a request to already existing intentions if they are "considered" close.
- Or by creating new intentions
|Zone 1||You will find in this part all the conversations not handled by the chatbot. You have the choice of either archiving certain requests or enriching your intentions with unrecognized sentences from users.|
|Zone 2||If you choose to enrich your intentions, a window opens with a drop-down menu containing all of the chatbot's intentions. You can select them by scrolling the page or by searching by name.|
Automatic or unsupervised learning¶
Now, when your Assistant is online, your users will ask various and varied questions which will allow you, if you wish, to automate other answers or to launch RPA (Workflow) according to certain requests. Or you probably already have millions or thousands of inquiries from your contact centers.
How do you make it easier for you to classify all these requests?
Prisme.ai's Machine Learning helps you categorize and group these requests. Prisme uses machine learning tools to create so-called cluster groups.
It is a decision support tool that allows the administrator to save time by optimizing and orienting the database of intentions according to user requests.
Just click on * suggest intentions * and let yourself be guided.