TAP
The Idea
This was before ChatGPT. Working in real-time captioning, I kept thinking about the years of transcripts just sitting there—what if we could extract information from them, or even “talk” to them? It could help with training materials and institutional resources, and unlock a window into the past. It felt like a treasure chest we didn’t have the key to yet.
What I Built
I didn’t know how to do any of this yet, so I partnered with a professor from GCCIS and we coded it together. I didn’t know about existing Python libraries or established NLP tooling, so once we landed on TF-IDF as the approach, we implemented it by hand.
What It Meant
This was my gateway into NLP and AI—not through a course or a tutorial, but through a problem I actually cared about, solved with someone who could teach me as we went. I had so much fun learning how all of this worked.
Sometimes the naive path—doing it by hand because you don’t know better yet—teaches you more than the efficient one would have.