Airwaves to Archives: Using ChatGPT to make WRMC recordings more accessible online
In a recent news post, Digital Projects and Archives Librarian Patrick Wallace shared how he used artificial intelligence (AI) tools to transcribe and summarize 423 historic WRMC radio recordings now available at the Internet Archive (go/ia).
As someone who specializes in describing archival materials, I wanted to make sure these transcribed and summarized recordings were easy to find.
My goal was to list all 423 WRMC recordings and their AI-generated summaries in ArchivesSpace, each linked to its audio file and transcript at the Internet Archive. Since we had already saved years of student effort using OpenAI Whisper and ChatGPT, I wanted to keep that momentum going so students could focus on more complex archival work.
While catalogers create records for books in Library Search (go/librarysearch), I create finding aids for archival materials in ArchivesSpace (go/aspace). I already had workflows for importing large spreadsheets into ArchivesSpace, but I didn’t know how to scrape metadata from the Internet Archive—and I’m not a coder! Still, as librarians and archivists, we’re good at finding information. Prompting ChatGPT to write scripts or walk me through repetitive tasks isn’t so different from crafting a thoughtful database search. Some comfort with the command line helps (and I don’t recommend experimenting on live collections), but this turned out to be a manageable project that closed the loop between description and access.
All 423 WRMC recordings are now listed individually in ArchivesSpace, complete with AI-generated summaries that improve keyword searching and direct links to the Internet Archive recordings.
Want to see one in action? Check out the March 1985 program, “This is the World of Technology,” in ArchivesSpace. You’ll find its AI summary in the “Additional Description” section and you can follow a link to the audio and transcript (“TEXT” under “Download Options”) at the Internet Archive.