post-template-default,single,single-post,postid-7787,single-format-video,theme-stockholm,qode-social-login-1.1.3,qode-restaurant-1.1.1,stockholm-core-1.2.1,woocommerce-no-js,select-theme-ver-5.2.1,ajax_fade,page_not_loaded,menu-animation-underline,side_area_slide_with_content,wpb-js-composer js-comp-ver-6.1,vc_responsive

Social network analysis and natural language understanding are what could be next

This is not a Rorschach, it’s 14 minutes of real-time on Twitter.

I’ve stumbled upon social network analysis and social network visualization, a subset. It’s a very different way to look at conversations on social media. The underlying theory has been in use for a few decades. Cops use it to identify and track gang organizations, governments have used it to take down terrorist networks etc.

#30_US hashtag for disinformation

#30_US hashtag for disinformation

Researchers and pathologists from the Chinese government and the World Health Organization are using social network analysis to track the outbreak of the Coronavirus. Follow the patterns of the people in the networks to determine who are the most influential; or in this case Patient Zero. In fact, here’s the link to a study on how researchers used network analysis during a previous coronavirus outbreak in the Nile Valley, 2012-2016


The image above is a snapshot of a recent Twitter conversation. On the night Trump tried to blow up the Middle East, a fake meme started spreading across Twitter. #30_US was a hashtag referring to a rumor that 30 US soldiers had been killed.


I used a tool called Gephi to monitor #30_US on Twitter in real-time that night. The image is a visualization of the conversation taken over 14 minutes. You can clearly see 3 primary communities and five smaller or looser ones. Using Gephi, you can drill down and look at the individual account, tweet, geo and other metadata.


I’m testing, and I stress I’M TESTING–but it looks like social network analysis can help communicators understand their actual social network. Gephi outputs a csv of each tweet and its metadata. I’m working on how to analyze these tweets using the IBM/Watson natural language understanding module. It would be interesting to see a conversation and analyze it for sentiment and emotion.


Yes…computers can now understand emotion in text.


For communicators, publicists, marketers and others from a wide range of industries could benefit greatly.  Imagine how you could respond if you knew the exact opinion leaders and the actual temperature of the issue?


The video above? I looked long and hard for a good vid on social network analysis. This one is the quickest and most informative after watching at least 5 pages of YouTube search results.