A Quest to Detect Novelty Using Deep Neural Nets
Dipannyta Nandi, Rohini Basak · 2020
In the present era, an abundance of data is found including images, videos and documents. Novelty has become an important matter to deal with. Here, focus is given on the novelty of documents specifically. In a sea of available documents how do we recognize the novel ones? With the help of deep neural networks an attempt is made to detect the novelty of documents. Detecting novelty of target documents follow the idea of binary classification. First, the similarity distance between a target document and some source document is calculated. Then a target document to source document mapping is done based on the distance value. This mapping is used as the input to a CNN model. We have also tried to analyse the effectiveness of Word Mover's Distance in detecting the novelty of the documents.