<!--
Purpose: The one SensAffect eVAD Lexicon — owner, survey, and site use
Last modified: 2026-09-09
By: Cursor Grok 4.6
Completeness: 100
-->

# SensAffect eVAD Lexicon

There is **one** SensAffect eVAD Lexicon. It is not a THRPY lexicon, a platform dummy, or a demo word list.

It is **smaller on purpose**: 131 authored feeling words, not a general-language VAD dump of thousands of tokens. Coverage is the words you can show a person. Accuracy is register plus orthogonal structure (VAD, tier, intensity), not raw token count.

Tier (general ↔ specific) and intensity (weak ↔ strong) **often travel together**, as on a Plutchik wheel. They are not the same: `angry` is general and fairly strong; `irritated` is weaker (mid, not a hub); `furious` is specific and strong.

## Source of truth

| Artefact | Role |
|----------|------|
| `lexicon/evad-lexicon-terms.mjs` | Authoring SSOT (tier, term, VAD) |
| `lexicon/evad-families.mjs` | Intensity families for that same list |
| `sensaffect-www/lexicon/evad-lexicon.json` | Published JSON |
| `sensaffect-www/design-system/evad-lexicon.js` | Live site (generated) |
| `sensaffect-www/lexicon/` | Public inspect page |

Rebuild after editing terms:

```bash
node scripts/build-evad-lexicon.mjs
```

## Dimensional and categorical (both stay)

Calvo & Kim 2012 keep two models, not one: a **dimensional** VAD vector from a normative vocabulary, and a **categorical** layer of named emotion centroids. Distance between a VAD point and those centroids is how a state is labelled. The 3-D model also drives visualization.

| Layer | SensAffect object |
|-------|-------------------|
| Dimensional / normative vocabulary | this lexicon (131 terms, VAD −1..1) |
| Categorical centroids | `VAD_CENTROIDS` in `cluster-bg.js` |
| Psychologically driven visualization | `EMOTION_THEMES` |

The lexicon does **not** replace centroids or themes. Dummy words (`weary`, `engaged`, `serenity`) stay out. Random or clinically damaging register (`cactus`, `trashbag`, `fart`, `dialysis`) stays out.

Categorical hub labels (Calvo & Kim, 1–9 rescaled to −1..1): `happy`, `angry`, `afraid`, `sad`, `neutral`.

Public page: https://sensaffect.com/lexicon/

Paper: https://doi.org/10.1111/j.1467-8640.2012.00456.x

## Hosts (not extra lexicons)

- **sensaffect.com** landing cluster and `/lexicon` use this list.
- **thrpy.chat** `/evad-survey.html` calibrates intensity order for this list.
- **THRPY Neural Map** may keep a working copy for the package build. Do not fork a second vocabulary.
- Phrase-sense analysis (`fine`, `I'm done`) is **lexical evidence**, not a second lexicon.

## Survey

https://thrpy.chat/evad-survey.html