A benchmark corpus to be used for measuring progress in statistical language
modeling. This has almost one billion words in the training data.
