Voice:
I am on the telephone with you. My actual voice is the data. Is the digitization of my voice not metadata? After all, natural voice =/= digitized voice. So it seems that digitized voice could be, to the NSA, metadata and therefore subject to capture.
I know the NSA has software/hardware to translate digitized voice to text. Suppose all phone conversations are converted to text. Then is that text therefore not data but metadata? Once again, it seems that the text of all phone conversations could be, to the NSA, metadata and therefore subject to capture.
Fax:
I send a fax to you. What's on the sheet of paper is data. But is the digitized and compressed sequence of bits sent to you not metadata? It certainly is not the same as the original data. So it seems that the digitized fax could be, to the NSA, metadata and therefore subject to capture.
Still, were that not so, the NSA has OCR that can read handwritten or typed text from a fax and convert it to text. Is the converted text not metadata? It is certainly different from the original data. So it seems that the converted text from the fax could be, to the NSA, metadata and therefore subject to capture.
E-mail:
I e-mail you. The text I see before me is data. Once I press the send key, it is compressed through various software/hardware for rapid transmission as it passes over the TCP/IP/phone network. Is not that compressed text also metadata? Therefore it seems that e-mail content could be, to the NSA, metadata and therefore subject to capture.
IOW once data is processed through any transformation whatsoever then it could be defined as "metadata". Again I don't know NSA's exact definition (used internally) of what metadata is. There may be many different definitions. But since NSA's representatives raised the use of the term w/o defining it (and then used the metaphor of a library to draw questioners off track) I believe that they should be questioned again about precisely what "metadata" means when and where. Furthermore the question of what precisely the NSA captures should be driven to ground thoroughly.
Personal belief:
NSA considers the product of any transformation of data to be metadata. The NSA captures all voice, fax, e-mail, chat, sms, etc. and maintains it in digitized form. Regardless of origin, all communications end up in text format (with pointers back to the original digitized form) which is then subjected to semantic/content analysis (scanning for naughty words), and optionally (under control of an analyst) to social network analysis and AI software that analyzes events, objects, actors, their intentions, and possible variations in interpretation of the text.