If it's been years, how do you know which of the 1700 are the relevant ones? Are you tagging or transcribing them? Seems like this would be a good use case for

If it's been years, how do you know which of the 1700 are the relevant ones? Are you tagging or transcribing them? Seems like this would be a good use case for even moderately accurate transcription of the audio files by software -- the way even moderately accurate OCR of a bag of scanned PDFs can provide the fodder for search tools like Spotlight to interactively surface approximately useful results.