Just had one of those "why didn't I try this sooner?" moments with Scholarcy. We all know its default keyword extraction is decent, but for deep technical papers or niche SRE topics, it can sometimes miss the really specific acronyms and tool names that matter.
I was reviewing a bunch of papers on eBPF-based networking and the summaries kept highlighting generic terms. Turns out, you can actually train it! There's a hidden(ish) feature where you can upload a custom glossary—a simple text file with one term per line. I threw in terms like `Cilium`, `kprobe`, `XDP`, `gRPC`, and `Teleport`. The next summary I ran picked up `XDP` and `Cilium` immediately, which completely changed the focus of the summary bullet points.
This is a game-changer for:
* **Literature reviews** – Seed it with your project's key terminology.
* **Keeping up in a niche** – Quickly surface papers that mention specific tools or protocols you care about.
* **Team onboarding docs** – Ensure summaries highlight your company's tech stack.
It's not a full ML training run, more of a priority nudge, but for technical deep dives, it makes a huge difference in relevance. Has anyone else played with this? Curious what other use cases you've found.
—Chris
K8s enthusiast