Repository
Python · JS · MITCore engine in Python, mirrored in JavaScript. Same word-list JSON drives both. Streamlit app for internal pipeline use; this static page for embed anywhere.
Live tool · Tools
Banned-Word Checker.
A browser tool that runs federal grant text through the leaked NSF 3-stage decision tree for flagged DEIA and Executive-Order language. Everything runs locally; nothing leaves your machine.
Need to go deeper? This is a quick, fixed-list check. For
configurable, topic-agnostic term discovery across your own word lists, use
Concordance ↗,
the lab's fuller desktop tool.
How the decision tree works
Faithful to the leaked NSF review diagram. The tree is hierarchical with early exit, which means a hit in the title produces a shorter report than a hit in the project description.
- Stage 1 — Title or Abstract: any flagged term here yields Category 3 and later stages are not scanned.
- Stage 2 — Project Summary: scanned only if Stage 1 was clean.
- Stage 3 — Project Description: scanned only if Stages 1 and 2 were clean.
- All clean: Category 1, which in the leaked process requires a reviewer comment explaining why no flagged language was found.
The word list merges (1) the leaked NSF source list with (2) a
community-compiled defensive list. Each term carries provenance —
leaked, existing, or both.
Privacy: all processing runs in your browser. The text you paste or the file you upload never leaves your computer. PDFs are parsed locally via PDF.js; DOCX via mammoth.js.
Loading word list…
Output
Upload a .docx or .pdf grant document
The tool attempts to split it by standard NSF section headers. You can edit any section in the paste view before running the check.
Open source · MIT
Inspect, fork, contribute.
Concordance
Desktop · robustThe fuller companion: a topic-agnostic term checker you point at any configurable word list, for deeper discovery than this fixed NSF list. AURA Lab in collaboration with SIM DAD LLC.