All2Vec

In this project, we try to authors of a anonymous paper in a double blind review setting. We consider a heterogenous graph and use Random walk based approach to learn the embeddings of various nodes in it like author, conference, venue, etc. We fit the embedding model with a daunting loss function and observe that when we use simple cosine similarity metric, we can identigy authors with a good accuracy. In addition to predicting the authors, we observed that we can also predict other attributes of an anonymized paper like venue etc.