Most chatbots rely on external APIs for the cool stuff such as natural language understanding (NLU) and disappoint because if and else conditionals fail at delivering good representations of our non linear human way to converse. Wouldn’t it be great if we could 1) take control of NLU and tweak it to better fit our needs and 2) really apply machine learning, extract patterns from real conversations, and handle dialogue in a decent manner? Well, we can, thanks to Rasa.ai. It’s open-source, it’s in Python, and it works.
Montréal-Python 68: Wysiwyg Xylophone https://montrealpython.org/2017/11/mp68/