Yes, I totally agree; it's not what I would qualify as open source.

Do you plan to release the training code along the research paper? What about the dataset?

In any case, it's very neat to have ML-based technique and lightweight model for such tasks!