Decode human behavior through code: A counter-intuitive approach

reyha (~reyha)


75

Votes

Description:

Human psychology has remained and continues to remain one of the most challenging areas of research as it aims to understand individual’s behavior and mind, including conscious and unconscious phenomena, as well as feeling and thought. The extent of impact social media has caused on the human mind is huge and perhaps, hard to imagine.

Thanks to python and it's brilliant capabilities to process natural language, we can now understand how social media is affecting our lives from a psychological perspective and if it is capable of changing our behaviors, our expressions, our sleeping patterns, or even emotions. From social posts, we can draw interesting conclusions about both men and women if we can comprehend what are the topics they are most interested in, what time of the day are they most and least active etc.

Core idea:

  1. Collect a dataset from Twitter (or any other social network) of the world's top 400 most influential women for the year 2013 and for the year 2018
  2. Train an NLP model and use this model to classify the collected data under various categories like education, religion, etc. and identify if the post is a concern, compliment, complaint etc.
  3. Perform a year-wise trend analysis to identify the topics they are most interested in and parameters like the most/least active time of the day, the most active/least active day of the week the average time spent on twitter per week/month etc.
  4. Carry out behavioral analysis by evaluating how the ways of expression, activity levels etc. have changed on social media over the last five years and what might have been the possible reasons for the same

Structure:

  1. 5-10 mins – Introduction and discussion ( algorithms and concepts being used )
  2. 10-20 mins – Code walkthrough followed by discussions on the results obtained (Please refer the core idea section for more details)
  3. Remaining time – Q/A or general discussion

Contents:

  1. An introduction to natural language processing - text normalization, n-grams, PoS tagging
  2. An introduction to deep learning - neural networks and neural language models (framework - keras)
  3. A brief discussion on the implementation of a sentiment classifier - Naive Bayes classifier/RNN classifier [HTML_REMOVED] If time permits, test out a few tweets to understand the working of the classifier
  4. Conclusion - how can the results help identify opinions, attitudes, emotional states & future scope (of the project)

Note: The entire talk will be a powerpoint based presentation along with illustrative code snippets

Prerequisites:

Python - Beginner/Intermediate [HTML_REMOVED] Machine Learning - Beginner [HTML_REMOVED] NLP/Deep learning techniques - Beginner [HTML_REMOVED] Keras/Tensorflow - Beginner [HTML_REMOVED][HTML_REMOVED] Basic familiarity with the following libraries/tools:[HTML_REMOVED] 1. numpy[HTML_REMOVED] 2. pandas[HTML_REMOVED] 3. matplotlib[HTML_REMOVED] 4. jupyter notebook[HTML_REMOVED]

Content URLs:

To be added soon!

Speaker Info:

The speaker of this talk is Reyha Verma. She is currently working as a data scientist at Sprinklr, Gurgaon. Since her organization is the world's best social media management platform, she spends most of her office and her personal time juggling between new, efficient deep learning models and tons of social media data.

She is an open-source enthusiast who has also previously been a mentor with Zulip, an open-source python based chat application for FOSS Outreachy program 2016 and has undertaken research projects at National Sun Yat-Sen University, Taiwan and Bhabha Atomic Research Center (BARC), Mumbai while pursuing her undergraduation at the National Institute of Technology, Srinagar.

Speaker Links:

LinkedIn - https://www.linkedin.com/in/reyhav [HTML_REMOVED] Github - https://github.com/reyha [HTML_REMOVED] Twitter - https://twitter.com/reyhav [HTML_REMOVED]

Id: 941
Section: Data science
Type: Talks
Target Audience: Intermediate
Last Updated: