Build a chatbot from scratch using open-source tools

Saikat Kumar Dey (~saikatkumardey)




This talk will show you how you could use open-source tools to build a chatbot with without handing your data over to third parties ( Facebook/Microsoft/Google )

You will learn:

  • how a chatbot works behind the scene
  • how to transfer knowledge to a chatbot
  • how to build a production-ready bot which you would own completely
  • how to use Socket.IO to make your chatbot communicate in real-time.


  • Introduction: motivations for building your bot in-house
  • Chatbot architecture.
  • Understanding the components of chatbot -

    • Natural Language Understanding (NLU)
      • Intent Classification
      • Entity Extraction
    • contextual dialogue management using Finite State Machines (FSM)
    • Response Generation
    • Knowledge Base
  • Open Source Tools to build your own chatbot:

    • Rasa.AI (having configurable backends like spacy/sklearn/mitie) - to build intent & entity extraction.
    • Transitions - to build a context manager using FSMs.
    • ChatterBot to generate automated response based on historical conversation.
    • Socket.IO - to build a real-time communication system.
  • Demo


  • Basic knowledge of NLP and Machine Learning.
  • Basic understanding of Finite State Machines.
  • Familiarity with Python.

Speaker Info:

Saikat Kumar Dey

Data Scientist at Endurance International Group. He has have been working on building a conversational assistant for Customer Support in my company. He has worked on customer segmentation & retention efforts in the past.

Parth Savjani

Data Scientist at Endurance International Group. He has 3+ years of experience in working with Data Infrastructure & predictive modelling on transaction data.

Section: Others
Type: Talks
Target Audience: Intermediate
Last Updated:

Hi, It would be nice if you add your slides before 10 Sept. It will help our team to review your proposal. Thanks

Rajat Saini (~rajataaron)

Hi, the slides have already been added under Content Url section. Thanks.

Saikat Kumar Dey (~saikatkumardey)

Login to add a new comment.