Using Pop Up Archive for oral history transcription

View Post on OUP Blog

By Samantha Snyder

After completing my first transcription process using Dragon NaturallySpeaking, I was asked to transcribe an interview using Pop Up Archive, an online platform for storing, transcribing, and searching audio content developed by the Public Radio Exchange (PRX). They explain the process in three steps:

1. You add any audio file.

2. We tag, index, and transcribe it automatically.

3. Your sound, in one place, searchable to the second.

I was assigned an interview from the University of Wisconsin collection of interviews regarding the Sterling Hall Bombing of 1970. This was a really interesting interview with a former undergraduate who was working at Sterling Hall on the night of the bombing. The uploading process is relatively simple, fitting well with the overall aesthetic of the website. You can either drag and drop files into the upload box or select one at a time. I chose to upload one file at a time, since I had four excerpts to work with. I later learned that if you choose to upload multiple files at once, they will merge into one large file. This, in my case, wouldn’t have been useful, as I wanted to test how long it took each excerpt to be transcribed.

The upload process was quick and painless, immediately giving me the option to add metadata, including the title, format, collection, images, and any tags. I decided to add a couple of tags (Sterling Hall, Sterling Hall Bombing, University of Wisconsin–Madison, Student protests, 1960s, 1970s) and a picture of Sterling Hall after the bombing. Without any prompting, Pop Up Archive began the transcription.

Here is the screen after the file was uploaded (Figure 1). The picture shows up as a little square next to the title. I find the screen to be a little too sparse. I understand that this website really focuses on being simplistic and easy to use, but I find the information to be oddly placed on the page.

The first excerpt was 4 minutes and 20 seconds long. I uploaded it at 3:20, and when I left my office for the day at 4:30, it was still transcribing. I received an email at 5:20 alerting me that the transcription was done. Whether this was exactly when the transcription finished, I’m not sure, but that seemed like an awfully long time for a short excerpt. Unfortunately, there isn’t a progress bar or anything along those lines to let you see how far along your transcription is.

"Pop Up Archive, Excerpt 1." Photo by Samantha Snyder.
“Figure 1, Sterling Hall Excerpt.” Photo by Samantha Snyder.

You can listen to the audio while the transcription is running in the background, so if this is an interview you haven’t heard, it’s a fun way to entertain yourself while waiting. I played around with the website to try and figure out what I could do while the file was being transcribed. I found that you can still edit the metadata while it is being transcribed and that there are additional metadata fields hidden away under a ‘more fields’ link, making it possible to add all kinds of information to the file.

When I got back to work the next day, I had the transcription ready and waiting for me to edit. I was incredibly impressed with the results. It did a nice job with the transcription, though I was able to spot some errors. The interviewee was soft-spoken and tended to run his words together slightly, so I was expecting there to be some editing that needed to be done. Besides the transcript, Pop Up Archive automatically adds suggestions for related topics that link to other interviews from other users. These are different than the user-created tags, because you cannot add more related topics, only delete suggestions that are not pertinent to your interview.

The transcription process had an easy learning curve and I felt like I was able to work efficiently. While you are editing the transcription, you can play the audio right along with it. You can stop and start the audio by pressing the Tab key and start the line over by using the command Shift and Tab. The one strange thing is that it splits the audio into separate lines, seemingly without any reasoning behind it, and there is no way to delete any of these line breaks, which sometimes were just a period or one word. You can add and delete words, but it will not recognize added words and move on to the next line if that is where it originally splits.

The software can differentiate between voices, but sometimes it recognizes too many voices. There were two interviewers and one interviewee in each piece, but the software recognized eight different voices in one excerpt. Fortunately, it’s an easy fix; you just assign speakers to the lines and it will fill down until it recognizes the next voice that you assigned. In the picture above, the letters “RW” are the initials of the interviewee.

I really enjoyed editing the transcription of the first excerpt and moved on to the second excerpt, which was a short 45 seconds. This took hardly any time at all. My third excerpt was seventeen minutes long, and I started the upload at 2:15 and received an alert email at 2:45. The fact that this took so much less time than the first excerpt makes me think that the platform may recognize voices after multiple files are uploaded with the same interviewer and interviewee.

"Pop Up Archive Transcription." Photo by Samantha Snyder.
“Pop Up Archive Transcription.” Photo by Samantha Snyder.

Though this excerpt was transcribed much faster, it had a larger amount of errors than the first and second excerpts. It also had lines that featured both the interviewer and interviewee speaking. Since there is no way to add new lines, I could only assign one speaker to the already created line. I had to be careful with this excerpt, since it was so long and had the bleed-over from interviewer to interviewee, there was a lot of room for mistakes. The picture above is of a portion of the transcription prior to editing. This section does not have any major mistakes, just some wording and grammar issues. To finish, I went through each excerpt to ensure the metadata was consistent, finished final edits of the transcripts, and confirmed they were all titled in a uniform fashion.

Since I was using this platform for research purposes, I did some searching to try and find exactly what software Pop Up Archive was using to do the transcribing and tagging. However, I had no such luck. This doesn’t worry me, but having a bit more information on how the transcription truly works would be helpful. I am slightly wary of putting faith in an online platform to transcribe and store all of your data, but they do give the option to download the audio and transcriptions which solves that, if you have adequate space to store the digital files. There is a possibility that the data could be lost or corrupt, but the same could happen when using servers and networks.

While there are things that could be improved, such as the ability to add and delete lines, a progress bar on the transcription process, and occasional grammar mistakes, this is a great program for oral history transcription. Transcribing these excerpts, which totaled about 23 minutes, took only about 20 hours to complete from start to finish, though most of this time was spent cutting the clips and uploading them to the service. With Dragon NaturallySpeaking, it took at least twice as long. I would highly recommend giving this platform a chance, though I cannot speak to the free transcription. I completed this process using the premium transcription.

Disclaimer: Pop Up Archive generously provided a free trial of their premium service for us to test out.

If you’ve tried transcription software, or other creative oral history methods, share your results with us in the comments below or on Twitter, Facebook, Tumblr, or Google+.

Image Credit: “Plaque on the south side of Sterling Hall on the University of Wisconsin-Madison campus” by JabberWock. CC BY SA 3.0 via Wikimedia Commons.