About the CSU Center for Public History + Digital Humanities

The CSU Center for Public History + Digital Humanities was established in 2007 to consolidate and manage various collaborative efforts and grant projects within Cleveland State University's History Department. The center's primary project, Cleveland Voices, began with faculty members collecting oral history interviews around 2002. Those interviews, along with others collected by CSU history students and local history educators, eventually grew into the Cleveland Regional Oral History Collection, which remains active and today consists of over 1,200 interviews. It is one of the largest geographically-specific oral history collections in the United States. These interviews cover a wide range of historical topics, both regional and national.

We spoke with Erin Bell, the project manager, where one of his responsibilities is cataloging and archiving the extensive oral history collection.

The Challenge

Initially, the Center relied on abstracts and minute-by-minute audio logs to document oral history interviews. While cost-effective, this method had significant limitations. Manual transcription was too expensive and time-consuming, and the audio logs provided only a rough outline of each interview, making it difficult for researchers to find specific details. This approach also excluded hearing-impaired researchers and hindered the overall discoverability of the collection's content.

Previous Solutions and Their Limitations

Over the years, we experimented with various AI-assisted transcription services and APIs, including Google Cloud and IBM's Watson. We even considered developing an open-source desktop app on top of a commercial API. Our primary goal was to enable interviewers to create transcripts before submitting their files for publication and to expedite the transcription of over a thousand older interviews.

In early 2020, we adopted a popular transcription solution, which offered an editing interface for the generated transcripts. We launched a campaign to cover the costs and hired graduate students to review and correct the transcripts. Despite some progress, the low quality of the uncorrected transcripts and the slow pace of progress required us to seek even more funding for professional human transcribers.

In early 2024, we resumed our search for a more efficient AI-assisted transcription service.

Discovering Alice

In our quest for a better solution, we discovered Alice. After a few test runs, we switched to Alice and haven't looked back. Alice's approach proved to be simpler, faster, more reliable, and more cost-effective than anything we had tried before. Alice may not have a built-in interface for editing transcripts but it really doesn't need one. The initial quality of the generated transcripts is so good that we can basically just read along while playing the audio at 2x speed, making only a few minor corrections here and there using any text editor. It's been about as close to perfect as one could reasonably hope.

Impact and Results

Since switching to Alice, we have experienced significant improvements in the transcription process. The high-quality transcripts generated by Alice have allowed us to make our oral history collection more accessible and searchable, enhancing the overall research experience. The transition to Alice has been seamless, and we are now able to process and publish transcripts at a much faster rate, ensuring that our extensive collection continues to grow and serve the needs of researchers and the public.