Chatbot
For a domain specific context
What is the motivation?
Considering the domain of students in a college or university, very often they need to contact customer care or student services.
People are used to receiving information instantly, however, a call to the customer care or an email to student services might experience no prompt response, in some cases taking days to answer simple questions.
What were the technologies used?
* Python
* Flask
* MongoDB
* Keras
* Rasa Framework
* Serverless Approach (front-end)
* Docker (back-end)
PRODUCT DETAILS
Main Functions
Used 20,000 different questions for training.
The bot is trained for more than 150 possible scenarios.
It can recognize more than 140 different intents.
It's now capable to recognize questions, extract intents and memorize contents.
Front-end
The purpose of this main site was to host the chatbot as a service, providing the initial information on how use and some social information about a college. The complete website designed as a single landing webpage to provide information at a glance.
RASA framework
Rasa was used to speed-up the build of a contextual assistant. An open-source well documented framework contributed to the final success. I took advantage of natural language understanding (NLU) and dialogue management to iterate fast and develop a multi-step conversations machine that remembers context and integrate business logic.
Docker
I choose an application containerization strategy to better isolate the application as well as to increase the scalability. A load balancer layer can also start or stop chat instances due to some automatic trigger events too.
About the code:
- Built with Python
- Ready to deploy to live website
Check the code:
nuRobot - Chatbot Source CodePublished: 10 JAN 2019