Lies, damned lies, and statistics
Reminds me a little of an excel sheet I whipped up to get me out of doing pointless work at my last job. The managers decided that every month, us engineers were to compile a report on where all the downtime was, and what had caused it, and how it impacted production. The operators logged all the production data in different spreadsheets on the network, so I set up SQL queries to find this info, add it all up for each area, categorize each cause, calculate efficiencies and list it out in a nice easy to read format. All I had to do was enter a date range to search and add some text to explain why the larger numbers were so much larger or smaller than the targets. The biggest issue was checking the input file to make sure the operators were using the correct format or miscategorizing
Compiling all of that information manually took at least 4 hours of tedious sorting, adding and subtracting, but I set up the program in about a day, with another day of tweaking. Saved me at least a day of sheer boredom every month.
It sounds to me like this would be something similar, only on a much larger scale. Keyword searches and statistics. Find out how many news articles or forum posts mention whatever you're interested in learning about and note whether there's more or less than last month. Cross reference whatever other points were being made in the same article. Categorize each article as being positive, negative or neutral based on wording. Interesting, but nothing too mysterious if you ask me.