Explore the process of building tools and frameworks for large-scale social media mining in this 46-minute talk by Dr. Juan M. Banda. Learn about the Social Media Mining Toolkit (SMMT) and its application in creating a massive COVID-19 Twitter dataset. Discover the challenges, lessons learned, and key decisions involved in developing and maintaining large-scale social media data gathering projects for NLP and machine learning research. Gain insights into the importance of standardization, scalable architecture, and iterative development in handling big data from social media platforms. Understand the benefits and cautions of using Twitter data, and explore the framework used for the COVID-19 data collection infrastructure.
Overview
Syllabus
Intro
Big data vs Large-scale?
KISS principle
Identify the scope of the problem
Avoid scope creep
Know your audience
Finding right tool for the job
Define a scalable architecture
5 Iterative development
Why social media?
Cautions about social-media data
Why Twitter?
Benefits of using Twitter
How do we harness such data?
The need for a specific tool
Beginning story (1)
So we needed to standardize this! (2)
In the end - lessons learned
For instant NLP uses
Defining a framework for data collection
Our COVID-19 infrastructure - under the hood (2)
Acknowledgments
Taught by
Elvis Saravia
