If you use the application please cite the paper. The input corpus is represented as two files: Cooccur.csv and Word2Id.csv. The first file follows the original libfm format (http://www.libfm.org) The folder data_pos contains the modeling of such example corpus using Part-of-Speech (POS) as additional contextual attribute, while the folder data_dependency contains the input data to train dependency-based embeddings. The Word2Id.csv file contains the symbols vocabulary, and looks like this: The first column contains the word form or the POS tag, the second column an unique identifier, the third column a meta information to distinguish words from POS tags (i.e., 0 for words, 2 for POS tag).