When technology companies get floated on the stock market, it prompts all kinds of analytical soul searching. Is Facebook really worth its $135 billion valuation? Is Zynga worth anything at all? Twitter is the latest Silicon Valley darling to offer the public a slice of its financial future. It is pricing its shares at $26 each, which values the company at more than $14 billion.
The company is pinning its hopes on boosted advertising revenues from its social media platform, but it also makes about $47 million a year from licensing its data to other firms. These companies can use the information to identify trends or monitor sentiment, for example. And it is not just big businesses that have an interest in this gold mine of data. Researchers also analyse it to gain insights into the way we behave. New Scientist has written about many of these over the years – here are five of the most exciting ones.
1. Tweet for the sick
You can use tweets to track the spread of disease. Adam Sadilek at the University of Rochester in New York and his colleagues used Twitter to follow the spread of flu virus in New York City. They used machine learning algorithms to search 4.4 million tweets for signs that people were feeling unwell. The system could differentiate between actual and metaphorical sickness, so “I’m sick of this traffic”, for example, wouldn’t register as illness. Combining this with location data, the team was able to see how the flu was travelling and predict when twitter users would fall ill. It could, perhaps, one day be used to warn people when they’re about to enter an area with a high infection rate.
2. The evolution of language
Twitter isn’t just good for telling you the latest news, it can also show how words are developing and spreading. Jacob Eisenstein of the Georgia Institute of Technology in Atlanta tracked the evolution of words. “Bruh”, a variation on “bro”, rose out of the south-eastern United States and made its way to California. “CTFU”, which stands for “cracking the fuck up”, emerged in Cleveland, Ohio, before spreading into Pennsylvania.
3. Beating the stock market
Can the fickle swing of Twitter conversation predict the billion-dollar jitters of the stock market? One hedge fund, Derwent Capital Markets, thought it could, but had to close its doors in May 2012. Twitter’s data can help predict stock market movement though. In October, The Wall Street Journal reported that subscribers to a company called Dataminr got an alert to take action 5 minutes before news of a shooting on Capitol Hill in Washington DC reached US TV. When the news hit the wider media it led to a 20 point drop in the Standard & Poor’s 500 financial index. Dataminr used an algorithmic assessment of Twitter to find the information fast.
4. Mapping America’s emotional state
Psychiatrists aren’t running machine learning on their patients’ Twitter profiles – yet. But large volumes of tweets can be used to make assertions about the happiness of large groups of people. A sentiment analysis run by Alan Mislove of Northeastern University in Boston measured every public tweet posted between September 2006 and August 2009, using a psychological word-rating system to identify happy or sad tweets. It turns out that the US west coast is happier than the east. Happiness peaks each Sunday morning, then dives to an all-week low on Thursday evenings.
5. Track food poisoning in restaurants
Social media messages can also tell us which places to eat might give our stomachs a nasty turn. Another system developed by Sadilek, called nEmesis, gathered 3.8 million New York tweets and ranked them for relevance to food poisoning. Messages containing words like “stomach” and “ill” were some of the key indicators that things weren’t right. A crowd of online workers then fell on the suspect tweets and ranked them according to likelihood that the tweeter had food poisoning. The human-generated results were used to automatically tag suspect tweets and show which restaurants might be best avoided.
Article on NewScientist.com