Pattern Recognition: ‘Fake news’ and other Machine Learning use cases
Here’s a puzzle for you:
3, 1, 2, 0, 1, -1, ?
240, 48, 12, 4, 2, ?
2, 5, 11, 17, 23, 31, ?
Find the pattern?
You may end up scratching your head a bit, but many of you will figure it out, without even pulling out a calculator. We humans are pretty good at pattern detection. And, we often use some form of pattern detection in our daily lives to help us solve problems or predict our next-best course of action.
Consider the store manager who observes sales trends to avoid having too little or too much product on the shelves. Or the baseball coach who studies how the team’s right-handed hitters fare against lefty pitchers, and then sets the lineup accordingly. Or the detective who looks for patterns at crime scenes or in a suspect’s behavior to build their case and prevent another crime.
We humans have always looked for patterns, but we’re limited by the amount of data our brains can “process” at any given moment. Analytic software and statistical tools help us build models to detect patterns and make predictions to extend our reach. Increasingly, there is enough data and complexity, that we can’t define a model capturing each input and output scenario up front.
That’s where machine learning (ML) comes in. With ML, the machine sets the model or learns and builds off a model rather than predetermining the inputs and outputs of each scenario. One of the primary applications of machine learning is pattern recognition, and in an age where countless inputs and billions of data points is all too common, we’re seeing ML taking off.
Thomson Reuters, a venerable source for news and information, demonstrates how companies are using ML to find patterns that would take us humans too long on our own or with traditional analytic tools.
In journalism, reporters are wired to look for patterns. The environmental reporter may look at health patterns in a local area to understand the effects of water contamination. A political reporter may watch for patterns in elections worldwide to identify changing political beliefs. A crime reporter may seek to find patterns across crimes or criminal behaviors to alert the public of new dangers.
In each case, the process is often the same. The journalist must keep a close eye on events, looking for activity that breaks from the usual patterns, and then validate its veracity by checking multiple sources and assessing each’s credibility.
But, like many of us, journalists have found there’s a lot more information to sort through than ever before. Social media, with Twitter in particular, is a growing source of leads, and spikes in words/phrases and conversations on Twitter can be a quick indicator of breaking news. But how does a journalist assess each conversation and its credibility in just minutes when there are thousands of sources? There’s a lot of false information, and, if not vetted properly, it can snowball into inaccurate news reports.
For Thomson Reuters journalists, ML is helping them discover breaking events, and do so before other news agencies. Reuters journalists can also focus on more in-depth investigations, as they’re not spinning their wheels trying to ascertain the credibility of the information or its sources.
ML makes it possible to capture and analyze huge numbers of Tweets, assess their credibility, and detect events, as they happen. It takes only 40 milliseconds for Thomson Reuters’ ML application, known as Reuters Tracer, to do this work. The initial intelligence is informed by journalists’ experience as the model is trained. But the application ultimately learns more complex patterns and its accuracy increases.
Read more about how Thomson Reuters is separating fact from rumors with Reuters Tracer and the Cloudera platform. Oh, and if you came up with 0, 2, and 41 earlier, you may not be able to keep up with the machines, but your pattern skills are sharp!
The post Pattern Recognition: ‘Fake news’ and other Machine Learning use cases appeared first on Cloudera VISION.