I don't know how they find those things out. It's just something that's too advanced for me. I think a guide for SOR1 was never created. One thing that I do know, judging from opening the ROM with Audacity (Signed 8-bit PCM, with a sampling rate of 7990hz), is that the voice clips appear to located near the end of the ROM.
I tried searching for several ROM addresses where the voices might be located, but all I get are 0.0 second lengths, Unknown lengths and voices that are at least over 1 second long.