Boost Surprising Discoveries for Online Health Information

In this project, we have created a framework for finding documents with surprising elements from a set of online health news documents. For that we have defined what is a surprise and developed a computational approach for finding surprising elements. Our model uses combination of machine learning classification algorithm and text analysis topic modelling for finding document with surprise element. Now a surprise can either be personalized or generalized. General surprise is something that violates the common knowledge of the entire society whereas personalized surprise is just for a particular person based on this person’s background knowledge. Personalized surprise is not necessarily surprising to society. We have trained our model in such a way that the document which we labelled as surprising, would be surprising for a group of persons(with similar background) instead of the whole society.