[Pycon] [new paper] "Akilesh Lakshminarayanan" - Building chatbots using Facebook Messenger API and Python-NLP packages

[info a pycon.it]

Title: Building chatbots using Facebook Messenger API and Python-NLP packages
Duration: 60 (includes Q&A)
Q&A Session: 15
Language: en
Type: Talk

Abstract: From helping you find the perfect pair of shoes to booking movie tickets, chatbots are reinventing the process of business-customer interaction. With more and more businesses focusing on customer retention by reducing user query response times, chatbots have come to the fore so much so that Facebook has now joined the army of messaging apps that expose APIs for developers to deploy their chatbots.

In the first segment of the talk, we will be taking you through the whole process of setting up your messenger bot service, starting from the authorisation, features and usage of the Messenger Platform API to deploying your application on a platform, like _Heroku_. 

After getting your bot up and running on Messenger, we add functionality to the bot by giving it Natural Language Processing capabilities. This will include text classification using _NLTK_, entity recognition, part-of-speech tagging using _spaCy_ language models that can learn over time. Then, we look at integrating Messenger frontend, NLP-modules and the backend database (using pandas) to generate appropriate responses to natural language user queries. 

Finally, we will be looking at different applications of conversational chatbots, what existing chatbots do, what future chatbots will do, and how your business can leverage them. Throughout the talk, I will be using an Airline customer service chatbot that I have programmed myself as an example, to help understanding the intricacies in designing functional chatbots.

This talk is aimed at novice to intermediate python programmers who are interested in Natural Language Processing & its applications. Some basic knowledge (like the existence!) of python-NLP packages would help.



Tags: [u'nltk', u'nlp', u'chat-bot', u'heroku', u'datascience', u'pydata']