scientific_authorship
Contains data used for authorship attribution classification experiments on scientific papers from ACL and EMNLP, as well as the source code for performing the classification experiments.
Data files
acl_pre2014.zip,emnlp_pre2014.zip,acl_201x.zip,emnlp_201x.zip: contains all of the articles used in the experiments, in .xml format, as extracted by the Grobid tool from the raw pdfs.vocabulary.pkl: contains a precomputed vocabulary of all words in the dataset, excluding rare words, as a serialized python dictionary.vocabulary_vectors.pkl: contains a precomputed list of embedding vectors for words in the vocabulary, as a serialized python dictionary with words as keys and embedding vectors as values. The embeddings are word2vec embeddings pretrained on Google News.
References
Full paper describing methodology and results can be found here: The Myth of Double-Blind Review Revisited: ACL vs. EMNLP, Cornelia Caragea, Ana Uban, Liviu P. Dinu