What Twitter Can Tell Us About Ebola
Gregory Alan Bolcer stashed this in Breaking Social
The last paragraph explains why Twitter has limited utility for sentiment analysis: Bots.
Between August 31 and October 15, 2014, Twitter users posted 5.4 million tweets that mentioned Ebola. That’s fewer than tweets referring to another widely covered, but localized, news event, the shooting of an unarmed black man by police in Ferguson, Mo; the hashtag #ferguson appeared in 7.8 million tweets in the nine days after the incident.
In tweets about Ebola, the most common topic discussed was possible treatments, from vaccines to homeopathic remedies. The phrase “cure for Ebola” was a more common term than “Ebola vaccine,” according to the startup.
Many tweets assigned blame for the illness, most commonly a national government. Tweets most often blamed the U.S., Liberia or Nigeria (the latter for allowing travel). After governments, the tweets most often blamed the airline industry, and then immigration.
Luminoso’s analysis is a lesson in the nature of viral social chatter. Twitter is effective at spreading valid information about current events in real time. But, as Luminoso’s data shows, it’s also effective at spreading unproven and baseless charges.
A portion of the tweets — roughly 80,000, or 1.5% — were devoted to conspiracy theories about the disease. The most prevalent was that Ebola was unleashed on Africans by Western governments as a form of population control intended to ensure the survival of whites. Another conspiracy theory was that Ebola was invented as a biological weapon by either a government or a terrorist organization.
Nine thousand tweets promoted the idea that the U.S. had developed an Ebola vaccine but withheld it so pharmaceutical companies could profit by selling treatments. This may seem like a relatively small number, but considering that the majority of tweets in the survey were news headlines –- i.e., “Breaking: First Ebola Death on US Soil” — the conspiracy theory theme is more salient, Luminoso said.
Luminoso scanned the text of tweets for verbal content and semantic structure and then performed sentiment analysis to infer the writer’s intention. Sentiment analysis of Twitter data is a growing area for business and governments. Luminoso’s core business is analyzing text from social media as well as customer service logs, emails and surveys to customers including Autodesk , General Mills , Sony and NASA.
Although Luminoso analyzed a large number of tweets, the sample likely included messages sent by automated bots and content taken from fake news sites. Such noise limits the utility of Twitter as a gauge of public opinion.