After a long enough time of knowing someone, one can get pretty accurate at predicting their friend's opinion. Now with so many people putting so many of their opinions online through blogs and social networking sites, it is possible to mine these posts and predict how a stranger would respond to something. Of course mining that much information is a very daunting task for any human, which is why computers are being taught to do this, and researchers at Concordia University have recently made a new tool for digging through the mass of data.
The greatest barriers to overcome for such software are related to informally-written language. Humans can have a tendency to digress from their point and throw in irrelevant information and even if the information is irrelevant, the sentences could be so poorly written that identifying their intent is unclear. Software has to be able to identify occurrences of these issues and respond appropriately, which BlogSum, the researchers' new tool, does with greater accuracy than its predecessors. Even when tested against human subjects, BlogSum was superior.
The ability to automatically analyze a blog post and then generate an accurate summary could prove very useful for a great many people. Researchers who use the Internet as a sample source of course would benefit, but also we may see such tools implemented in search engines, to bring up more appropriate results when requested.