Week 3: Transcription Bumps and AI

    Hello, my name is Hiram Davila. I am in my second year at UCF in the History Master's program, and I am extremely excited to share some information about my internship this semester. My current research interests regard the early modern Spanish Empire, particularly the Spanish Inquisition. This blog will be acting as my journal week-by-week as I intern with the city of Orlando and Greenwood Cemetery throughout the Fall 2024 semester.

    This week, I did much of the same as last week. A few of the problems I have ran into with the this transcript include using the correct pronunciations, such as when someone says "uh" in the middle of the sentence, making sure that proper nouns are spelled correctly and doing quick additional research outside of the oral history to make sure they are accurate when I can, as well as differentiating when to add verbs. For example, if X person is talking and Y person laughs, you are supposed to put Y: [laughs] and continue X's paragraph. However, I am running into some spots where the interviewer very slightly laughs, or lets out a soft chuckle, where if you are only listening to the audio you might not hear it, and I am unsure if I should put those into the transcripts or not. There were other challenges as well, but I would encourage you to look at my previous 2023 Spring RICHES blog between weeks 2 and 4 for more of the same experience I have had with transcripts.

    Something different that I wanted to add this week was that I have ran into AI for the first time on this project. Taking a step back to my time at RICHES in 2023, all the oral histories that I watched were YouTube videos with no captions available and I had to just us my ears to make the transcription. In this project, the oral history A/V files are uploaded via Microsoft Sharepoint. For this transcript in particular, Sharepoint has auto-generated some of the transcript for me. While it is still taking me some time to go through the video because I do not fully trust the AI, and its not very good at punctuation or differentiating words when a person might mumble, I would say it has about a 65% accuracy rate. This transcript that I am working on now is much longer than the one I worked on in RICHES, and so the AI generation is making this process a bit more faster.

    I think this is an important point to bring up because historians and the world at large are moving closer and closer to becoming more reliant on AI. AI has exploded in just the past 2 years alone and are becoming increasingly more versatile. For me, the reason why I think it was important to talk about the AI-generated transcript is because I have a very negative connotation of AI in academia. I believe that it has been used by a lot of students the past few years both at the graduate and undergraduate level, as well as in K-12. My little sister even uses it to assist with some of her assignments as a 10th grader (though I was quick to inform her about how to use AI properly to assist with assignments, and not to just let AI do it for you). I see it as a huge threat to academic writing, thought, and impacts the creativity of people. If you are just using a computer to think for you, you are not developing vital critical thinking skills needed both in use in academia and the real world. Sharepoint's AI, however, has at least opened my eyes to beneficial ways AI can be used without much harm. It has made this transcription much easier to make as I continue working on it and overall I think is a helpful tool that will help me be more efficient in general with the other transcripts that I will be working on in this internship in the future.