How we calculate this score
We use a free, open-source semantic quantifier Python package called
NRCLex (version 4.x) that you can install yourself with
pip install NRCLex
or browse on GitHub at
github.com/metalcorebear/NRCLex.
It is built on the public
NRC Emotion Lexicon
(word lists labeled with emotions such as fear, anger, trust, and sadness).
We feed each Wire report or transcript into that library, count how many words match the
“fear” (threat-language) list, and report the result as
hits per 1,000 words
so a short daily Wire and a long briefing can be compared fairly.
We then add a modest boost from an S2-specific threat term list
(words like attack, strike, kill, bombing — military wording the general NRC list often
under-weights), scaled so it nudges the gauge without replacing NRCLex.
The needle on our meter is that blended number; the center mark is the long-run average across
hundreds of past briefings (~34.3 on the blended scale), and the left/right ticks are about one
standard deviation below and above that average (~±11.9), with a “high” band near mean + 1.5σ
(~52). Those baselines were recomputed from 500+ Wire/Intel texts using the same blend.
It is a language-density gauge, not a clinical or psychological diagnosis.
In plain steps
- Take the text of the report (cleaned of headers where needed).
- Run it through NRCLex, which tags words against the NRC emotion lists.
- Count fear-tagged hits, divide by total word count, multiply by 1,000 (pure NRC score).
- Count S2 threat-lexicon hits the same way and add a small fraction of that density.
- Compare the blended score to our historical average and spread (standard deviation) for the meter.
What it is not
It does not know sarcasm, negation (“not a threat”), or real-world events—only whether the wording overlaps a public emotion lexicon. Military and conflict vocabulary can raise the score even in calm analysis. Treat it as a relative “threat-language density” index for this corpus, not a claim about anyone’s mental state.