Tag Archives: audio

Two Years In: The Finish Line Approaches for Digitizing Behind the Veil

Behind the Veil Digitization intern Sarah Waugh and Digital Collections intern Kristina Zapfe’s efforts over the past year have focused on quality control of interviews transcribed by Rev.com. This post was authored by Sarah Waugh and Kristina Zapfe.

Introduction

The Digital Production Center (DPC) is proud to announce that we have reached a milestone in our work on Documenting African American Life in the Jim Crow South: Digital Access to the Behind the Veil Project Archive. We have completed digitization and are over halfway through our quality control of the audio transcripts! The project, funded by the National Endowment for the Humanities, will expand the Behind the Veil (BTV) digital collection, currently 410 audio files, to include the newly digitized copies of the original master recordings, photographic materials, and supplementary project files.

The collection derives from Behind the Veil: Documenting African-American Life in the Jim Crow South. This was an oral history project headed by Duke University’s Center for Documentary Studies from 1993 to 1995 and is currently housed in the David M. Rubenstein Rare Book and Manuscript Library and curated by the John Hope Franklin Research Center for African and African American History and Culture. The BTV collection documented and preserved the memory of African Americans who lived in the South from the 1890s to the 1950s, resulting in a culturally-significant and extensive multimedia collection. 

As interns, our work focused on ordering transcripts from Rev.com and performing quality control on transcripts for the digitized oral histories. July 2023 marked our arrival at the halfway point of completing the oral history transcript quality control process. At the time of writing, we’ve checked 1727 of 2876 files after a year of initial planning and hard work. With over 1,666 hours worth of audio files to complete, 3 interns and 7 student workers in the DPC contributed 849 combined hours to oral history transcript quality control so far. Because of their scope, transcription and quality control are the last pieces of the digitization puzzle before the collection moves on to be ingested and published in the Duke Digital Repository

We are approaching the home stretch with the deadline for transcript quality control coming in December 2023, and the collection scheduled to launch in 2024. With that goal approaching, here is what we’ve completed and what remains to be done.

Digitization Progress

A graphic showing the statistics of the Behind the Veil project alongside a tablet device of the Duke Libraries Behind the Veil webpage. Under Audio, the text reads: 1,577 tapes, 2.876 audio files, and 2,709 files transcribed. Under the word Admin Files, it reads: 27,737 image files, under the Prints and Negatives heading it reads, 1,294 image files. On the right side of the graphic, headings for Video, Project Records, and Photo Slides are here and under Video it reads 14 tapes, 14 video files, under Project Records, it reads 9,328 image files, and under Photo Slides, it reads 2,920 image files.

As the graphic above indicates, the BTV digitization project consists of many different media like audio, video, prints, negatives, slides, administrative and project related documents that tell a fuller story of this endeavor. With these formats digitized, we look forward to finishing quality control and preparing the files for handoff to members of the Digital Collections and Curation Services department for ingest, metadata application, and launch for public access in 2024. We plan to send all 2876 audio files to Rev.com service by the end of August and to perform quality control on all those transcripts by December 2023.

Developing the Transcription Quality Control Process

With 2876 files to check within 19 months, the cross-departmental BTV team developed a process to perform quality control as efficiently as possible without sacrificing accuracy, accessibility, and our commitment to our stakeholders. We made our decisions based on how we thought BTV interviewers and narrators would want their speech represented as text. Our choices in creating our quality control workflow began with Columbia University’s Oral History Transcription Style Guide and from that resource, we developed a workflow that made sense for our team and the project. 

Some voices were difficult to transcribe due to issues with the original recording, such as a microphone being placed too far away from a speaker, the interference of background noise, or mistakes with the tape. Since we did not have the resources to listen to entire interviews and check for every single mistake, we developed what we called the “spot-check” process of checking these interviews. Given the BTV project’s original ethos and the history of marginalized people in archives, the team decided to prioritize making sure race-related language met our standards across every single interview.

A few decisions on standards were quick and unanimous—such as not transcribing speech phonetically. With that, we avoided pitfalls from older oral histories of African Americans, like the WPA’s famous “Slave Narratives” project, that interviewed formerly-enslaved people, but often transcribed their words in non-standard phonetic spellings. Some narrators in the BTV project who may have been familiar with the WPA transcripts specifically requested the BTV project team not to use phonetic spelling. 

Other choices took more discussion: we agreed on capitalizing “Black” when describing race, but we had to decide whether to capitalize other racial terms, including “White” and antiquated designations like “Colored.” Ultimately, we decided to capitalize all racial terms (with the exception of slurs). The team did not want users to make distinctions between lower and uppercase terms if we did not choose to capitalize them all. Maintaining consistency with capitalization would provide clarity and align with BTV values of equality between all races.

Using a spot-check process where we use Rev’s find-and-replace feature to standardize our top priorities saved us time to improve the transcripts in other ways. For instance, we also try to find and correct proper nouns like street names or names of important people in our narrators’ communities, allowing users to make connections in their research. We corrected mistakes with phrases used mainly in the past or that are very specific to certain regions, such as calling a dance hall a “Piccolo joint” from an early jukebox brand name. We also listened to instances where the transcriptionist could not hear or understand a phrase and marked it as “indistinct,” so we can add in the dialogue later (assuming we are able to decipher what was said). 

While we developed these methods to increase the pace of our quality control process, one of the biggest improvements came from working with Rev. If we were able to attain more accurate transcripts, our quality control process would be more efficient. Luckily, Rev’s suite of services provided us this option without straying too far from our transcription budget.

Improving Accuracy with Southern Accents Specialists

When deciding on what would be the best speech-to-text option for our project’s needs, we elected to order Transcript Services from Rev, rather than their Caption Services. This decision hinged on the fact that the Transcript Services option is their only service that allows us to request Rev transcriptionists who specialize in Southern accents. Many people who were interviewed for Behind the Veil spoke with Southern accents that varied in strength and dialect. We found that the Southern accent expertise of the specialists had a significant impact on the accuracy of the transcripts we received from Rev. 

This improvement in transcript quality has made a substantial difference in the time we spend on quality control for each interview: on average, it only takes us about 48 seconds of work for every 60 seconds of audio we check. We appreciated Rev’s offering of Southern accent specialists enough that we chose that service, even though it meant that we had to then convert their text file format output to the WebVTT file format for enhanced accessibility in the Duke Digital Repository.   

Optimizing Accessibility with WebVTT File Format

The WebVTT file format provides visual tracking that coordinates the audio with the written transcript. This improvement in user experience and accessibility justified converting the interview transcripts to WebVTT format. Below is a visual of the WebVTT format in our existing BTV collection in the DDR. Click here to listen to the audio recording.

We have been collaborating with developer Sean Aery to convert transcript text files to WebVTT files so they will display properly in the Duke Digital Repository. He explained the conversion process that occurs after we hand off the transcripts in text file format.

“The .txt transcripts we received from the vendor are primarily formatted to be easy for people to read. However, they are structured well enough to be machine-readable as well. I created a script to batch-convert the files into standard WebVTT captions with long text cues. In WebVTT form, the caption files play nicely with our existing audiovisual features in the Duke Digital Repository, including an interactive transcript viewer, and PDF exports.”  Sean Aery, Digital Projects Developer, Duke University Libraries

Before conversion, we complete one more round of quality control using the spot-checking process. We have even referred to other components of the Behind the Veil collection (Administrative and Project Files Administrative Files) to cross-reference any alterations to metadata for accuracy.

Connecting the Local and Larger Community

Throughout the project, team members have been working on outreach. One big accomplishment by project PI John Gartrell and former BTV outreach intern Brianna McGruder was Behind the Veil at 30: Reflections on Chronicling African American Life in the Jim Crow South.” This 2-day virtual conference convened former BTV interviewers and current scholars of the BTV collection to discuss their work and the impact that this collection had on their research. 

We also recently presented at the Triangle Research Libraries Network annual meeting, where our presentation overlapped with some of what you’ve just read in this post. It was exciting to share our work publicly for the first time and answer questions from library staff across the region. We will also be presenting a poster about our BTV experience at the upcoming North Carolina Library Association conference in Winston-Salem in October.

A image of two people standing a podium with a screen behind them. Four people in the front row look out at them.
Sarah Waugh and Kristina Zapfe presenting at the 2023 TRLN Annual Conference.

As we’ve hoped to convey, this project heavily relies on collaboration from many library departments and external vendors, and there are more contributors than we can thoroughly include in this post. Behind the Veil is a large-scale and high-profile project that has impacted many people over its 30-year history, and this newest iteration of digital accessibility seeks to expand the reach of this collection. Two years on, we’ve built on the work of the many professionals who have come before us to create and develop Behind the Veil. We are honored to be part of this rewarding process. Look for more BTV stories when we cross the finish line in 2024. 

Mythical Beasts of Audio

Gear. Kit. Hardware. Rig. Equipment.

In the audio world, we take our tools seriously, sometimes to an unhealthy and obsessive degree. We give them pet names, endow them with human qualities, and imbue them with magical powers. In this context, it’s not really strange that a manufacturer of professional audio interfaces would call themselves “Mark of the Unicorn.”

Here at the Digital Production Center, we recently upgraded our audio interface to a MOTU 896 mk3 from an ancient (in tech years) Edirol UA-101. The audio interface, which converts analog signals to digital and vice-versa, is the heart of any computer-based audio system. It controls all of the routing from the analog sources (mostly cassette and open reel tape decks in our case) to the computer workstation and the audio recording/editing software. If the audio interface isn’t seamlessly performing analog to digital conversion at archival standards, we have no hope of fulfilling our mission of creating high-quality digital surrogates of library A/V materials.

Edirol UA-101
The Edirol enjoying its retirement with some other pieces of kit

While the Edirol served us well from the very beginning of the Library’s forays into audio digitization, it had recently begun to cause issues resulting in crashes, restarts, and lost work. Given that the Edirol is over 10 years old and has been discontinued, it is expected that it would eventually fail to keep up with continued OS and software updates. After re-assessing our needs and doing a bit of research, we settled on the MOTU 896 mk3 as its replacement. The 896 had the input, output, and sync options we needed along with plenty of other bells and whistles.

I’ve been using the MOTU for several weeks now, and here are some things that I’m liking about it:

  • Easy installation of drivers
  • Designed to fit into standard audio rack
  • Choice of USB or Firewire connection to PC workstation
  • Good visual feedback on audio levels, sample rate, etc. via LED meters on front panel
  • Clarity and definition of sound
MOTU 896mk3
The MOTU sitting atop the audio tower

I haven’t had a chance to explore all of the additional features of the MOTU yet, but so far it has lived up to expectations and improved our digitization workflow. However, in a production environment such as ours, each piece of equipment needs to be a workhorse that can perform its function day in and day out as we work our way through the vaults. Only time can tell if the Mark of the Unicorn will be elevated to the pantheon of gear that its whimsical name suggests!

Future Retro: Images of Sound Technology in the 1960s Duke Chronicle

Many of my Bitstreams posts have featured old-school audio formats (wax cylinder, cassette and open reel tape, Minidisc) and discussed how we go about digitizing these obsolete media to bring them to present-day library users at the click of a mouse.  In this post, I will take a different tack and show how this sound technology was represented and marketed during its heyday.  The images used here are taken from one of our very own digital collections–the Duke Chronicle of the 1960s.

The Record Bar

Students of that era would have primarily listened to music on vinyl records purchased directly from a local retailer.  The advertisement  above boasts of “complete stocks, latest releases, finest variety” with sale albums going for as little as $2.98 apiece.  This is a far cry from the current music industry landscape where people consume most of their media via instant download and streaming from iTunes or Spotify and find new artists and songs via blogs, Youtube videos, or social media.  The curious listener of the 1960’s may have instead discovered a new band though word of mouth, radio, or print advertising.  If they were lucky, the local record shop would have the LP in stock and they could bring it home to play on their hi-fi phonograph (like the one shown below).  Notice that this small “portable” model takes up nearly the whole tabletop.

Phonograph

The Moon

Duke students of the 1960s would have also used magnetic tape-based media for recording and playing back sound.  The advertisement above uses Space Age imagery and claims that the recorder (“small enough to fit in the palm of your hand”) was used by astronauts on lunar missions.  Other advertisements suggest more grounded uses for the technology:  recording classroom lectures, practicing public speaking, improving foreign language comprehension and pronunciation, and “adding fun to parties, hayrides, and trips.”

Tape Your Notes

Add a Track

Creative uses of the technology are also suggested.  The “Add-A-Track” system allows you to record multiple layers of sound to create your own unique spoken word or musical composition.  You can even use your tape machine to record a special message for your Valentine (“the next best thing to you personally”).  Amplifier kits are also available for the ambitious electronics do-it-yourselfer to build at home.

Tell Her With Tape

Amplifier Kit

These newspaper ads demonstrate just how much audio technology and our relationship to it have changed over the past 50 years.  Everything is smaller, faster, and more “connected” now.  Despite these seismic shifts, one thing hasn’t changed.  As the following ad shows, the banjo never goes out of style.

Banjo

 

What’s DAT Sound?

My recent posts have touched on endangered analog audio formats (open reel tape and compact cassette) and the challenges involved in digitizing and preserving them.  For this installment, we’ll enter the dawn of the digital and Internet age and take a look at the first widely available consumer digital audio format:  the DAT (Digital Audio Tape).

IMG_0016

The DAT was developed by consumer electronics juggernaut Sony and introduced to the public in 1987.  While similar in appearance to the familiar cassette and also utilizing magnetic tape, the DAT was slightly smaller and only recorded on one “side.”  It boasted lossless digital encoding at 16 bits and variable sampling rates maxing out at 48 kHz–better than the 44.1 kHz offered by Compact Discs.  During the window of time before affordable hard disk recording (roughly, the 1990s), the DAT ruled the world of digital audio.

The format was quickly adopted by the music recording industry, allowing for a fully digital signal path through the recording, mixing, and mastering stages of CD production.  Due to its portability and sound quality, DAT was also enthusiastically embraced by field recordists, oral historians & interviewers, and live music recordists (AKA “tapers”):

tapers[Conway, Michael A., “Deadheads in the Taper’s section at an outside venue,” Grateful Dead Archive Online, accessed October 10, 2014, http://www.gdao.org/items/show/834556.]

 

However, the format never caught on with the public at large, partially due to the cost of the players and the fact that few albums of commercial music were issued on DAT [bonus trivia question:  what was the first popular music recording to be released on DAT?  see below for answer].  In fact, the recording industry actively sought to suppress public availability of the format, believing that the ability to make perfect digital copies of CDs would lead to widespread piracy and bootlegging of their product.  The Recording Industry Association of America (RIAA) lobbied against the DAT format and attempted to impose restrictions and copyright detection technology on the players.  Once again (much like the earlier brouhaha over cassette tapes and subsequent battle over mp3’s and file sharing) “home taping” was supposedly killing music.

By the turn of the millennium, CD burning technology had become fairly ubiquitous and hard disk recording was becoming more affordable and portable.  The DAT format slowly faded into obscurity, and in 2005, Sony discontinued production of DAT players.

In 2014, we are left with a decade’s worth of primary source audio tape (oral histories, interviews, concert and event recordings) that is quickly degrading and may soon be unsalvageable.  The playback decks (and parts for them) are no longer produced and there are few technicians with the knowledge or will to repair and maintain them.  The short-term answer to these problems is to begin stockpiling working DAT machines and doing the slow work of digitizing and archiving the tapes one by one.  For example, the Libraries’ Jazz Loft Project Records collection consisted mainly of DAT tapes, and now exists as digital files accessible through the online finding aid:  http://library.duke.edu/rubenstein/findingaids/jazzloftproject/.  A long-term approach calls for a survey of library collections to identify the number and condition of DAT tapes, and then for prioritization of these items as it may be logistically impossible to digitize them all.

And now, the answer to our trivia question:  in May 1988, post-punk icons Wire released The Ideal Copy on DAT, making it the first popular music recording to be issued on the new format.