RefWeb is aimed at helping you perform a complete literature survey through forward and backward chaining of references. The idea is to start off with a seed paper and then resolve all forward and backward references. Forward references is avaible on google scholar. Backward references can be extracted from a pdf of the paper. All resolved references are pushed to a queue. For each item in the queue the user will be asked if the paper is relevant. If that is teh case, it should be catagorised and saved. It's references should in turn then also be pushed to the queue.
Everything. I started with an attempt to extract references from PDF's. A citation parser is probably required so that the citations can be interpreted and the associated PDF can be found.