citracer traces keyword citations through papers

This is a genuinely useful niche tool for literature review archaeology, especially when you remember a concept but not its origin. The tradeoff is that its quality will live or die on parser accuracy and citation coverage, so it will shine on ML/CS papers and get shakier elsewhere.

• –The forward mode is the interesting part, because it ties keyword occurrences to nearby references instead of just building a generic bibliography network
• –The reverse mode is pragmatic, not magical, since it skips PDF downloads and leans on Semantic Scholar's citation-context metadata
• –GROBID dependence is a real constraint, but also the right one for this problem if the target domain is paper-heavy and citation-rich
• –The project fits the "research workflow infrastructure" lane more than a flashy end-user app, which makes it appealing to power users and contributors
• –The rough edges the author calls out are believable, but the scope is narrow enough that a focused community could improve it quickly