The corpusEN.bin file is the training data provided by OpenNLP which I used to split sentences (http://opennlp.sourceforge.net/models-1.5/). It's not the training data used for summarization.