The call-to-action [0] in the submitted article is meant more for serendipitous crowdsourcing, so the image/scan is presumably enough for that usecase. The actual dump (i.e. "tranche") of memos is hosted on DocumentCloud, which has the underlying OCR data per page, and is searchable:

https://www.documentcloud.org/documents/4357755-11-L-0559-Fi...

You can directly download the PDF (913 pages) here:

https://assets.documentcloud.org/documents/4357755/11-L-0559...

[0] https://www.muckrock.com/assignment/help-explore-donald-rums...