Research
What 50,000 abstracts taught us about how researchers describe their work.
Patterns in how fields phrase novelty — and what that means for matching.
By the Ravo teamFeb 22, 2026 · 7 min read
We analyzed 50,000 abstracts across five fields to understand how researchers signal what's new. The vocabulary of novelty turns out to be strikingly field-specific.
Novelty has a dialect
Machine-learning abstracts lead with benchmarks and deltas; neuroscience leads with mechanism; computational biology leads with scale and release. A matcher that ignores these dialects over-weights surface keywords and misses the actual contribution.
So our scope matching is now field-aware — it reads the same abstract differently depending on where it's likely to be submitted. That single change moved our top-3 accuracy more than any model swap did.