Replication Data for: Experiments with Learning Graphical Models on Text

This dataset contains the binarised document collections in .arff format to replicate the Experiments with Learning Graphical Models on Text. Every zip file contains a train a test folder. In each folder, there also are two set of files corresponding to the version with 500 words or 2000 words vocabulary. For every set, there are several files corresponding to the different training/testing splits and to the 5 repetitions per experiment. The test folders also contain files that start with WORD2PREDICT, which contain the words in the test set to prediction for the prediction task.