Researchers Mine Social Media for New Drug Termsby Chris Carroll | Photo Illustration by John T. Consoli and Gabriela Hernandez
Acting on a tip from an epidemiologist in Florida, researchers at the University of Maryland Center for Substance Abuse Research (CESAR) combed through Twitter posts in early 2015 in hopes of learning about a mysterious synthetic street drug known as flakka.
Just as they discovered its apparent origin—the Dominican Republic—and informed other researchers, flakka exploded in the news, with users suffering from violent psychotic episodes and dying from overdoses. Could things have been different if law enforcement and public health authorities had known earlier what drug users were talking—and tweeting—about?
Today, CESAR is working with experts in linguistics and computing at UMD’s Center for Advanced Study of Language (CASL) to use social media and big data analysis to develop a lightning-fast, automated method to uncover novel drug terms. It’s part of CESAR’s mission to develop and run the National Drug Early Warning System (NDEWS) for the National Institute on Drug Abuse.
“There’s a predictive element to this” that might help authorities avoid being taken by surprise by new drugs, says Eric Wish, CESAR director and NDEWS principal investigator.
Researchers at CASL use computer algorithms to analyze a sampling of hundreds of millions of North American tweets each month. The algorithms key in on known drug-related terms supplied by CESAR, and scoop up contextually similar unknown words, which are sent to CESAR for study.
Linguists have observed that groups discussing taboo subjects have quickly evolving vocabularies, says Claudia Brugman, technical director of language and social systems at CASL. Human analysts poring over tweets, in addition to being incomparably slower than computers, could have gaps in knowledge that would cause them to overlook new terms.
“If you want to learn how people are talking about a certain drug and just do a search for ecstasy, for instance, you’re going to miss a lot,” Brugman says.
Although the collaboration is in its infancy, the model has performed strongly so far in tests searching for new terms related to use of cannabis and MDMA, or ecstasy, the researchers say. Future development could include other social media platforms as the investigators cast a broader net.
“It remains to be seen how willing people will be to tweet about their drug use,” Wish says. “But if you want to study something, you need to know what to call it.”
Leave a Reply
* indicates a required field