DISQUS

NickAtFlare · 4 months ago

What is the source of your PhD magic you used to analyse the Twitter reaction? I'm intrigued to learn more about the process.

Melyssa Plunkett · 4 months ago

NickAtFlare,
Our PHD magic is our propietary IP, what we call our "secret sauce". The algorithm was developed at Harvards Institute for Quantative Social Science by one of our founders, Dr Gary King. Here is a link to the whitepaper that described the algorithm in detail. Email me at mog@crimsonhexagon.com if you'd like more details about the process.

Melyssa Plunkett · 4 months ago

correction - mpg@crimsonhexagon.com

Fox Krieger · 4 months ago

The only problem with doing stats (of any type, but especially based on counting commonly used phrases) is that many many Twitter spammers usually have tons of accounts where they will tweet the exact same message to all of them...

Do a search for a trending keyword and watch the results for a while (either using twitter's own search function or something like twitterfall) and you'll begin to see the EXACT same wording appear again and again in tweets from different 'people'. Further investigation will usually reveal those tweets are coming from 'sock puppet' or fake accounts.

Melyssa Plunkett · 4 months ago

Fox Krieger,
You are absolutely correct: keyword matching is an ineffective method for analyzing social media content. Our technology does not use keywords for the analysis. It uses keywords to collect the corpus of data to be analyzed but then leverages our statistical algorithm to analyze the data, recognizing and quantiying opinions and themes. We can also use this algorithm to filter out spam, again recognizing patterns and themes common across the spam content.