pdfOCR: Who provided TESS_DATA_DIRECTORY?

The tessdata directory should be retrieved by you, as there are multiple repositories with different training models. A great place to start is Tesseract's own GitHub repo, where you can find the tessdata (standard model) and the tessdata_best (slower, but higher accuracy) data files. 

After downloading the training models, you just need to point to them with pdfOCR, by using the appropriate method (Java/.NET).