Java Deep Learning Cookbook

How it works...

In step 1, the writeWord2VecModel() method saves the Word2Vec model into a compressed ZIP file and sends it to the output stream. It saves the full model, including Syn0 and Syn1. The Syn0 is the array that holds raw word vectors and is a projection layer that can convert one-hot encoding of a word into a dense embedding vector of the right dimension. The Syn1 array represents the model's internal hidden weights to process the input/output.

In step 2, the readWord2VecModel()method loads the models that are in the following format:

• Binary model, either compressed or not compressed
• Popular CSV/Word2Vec text format
• DL4J compressed format

Note that only weights will be loaded by this method.