Try the PV Peak/Valley Explorer
🔗 PV Radar (Beta) Space: https://huggingface.co/spaces/jsisonou/narrative-engine-pv-radar-beta
Use this dataset’s sample files to test:

Curve Mode: upload book_curve.scene.csv → Run

Text Mode: paste one scene per line → Run
You’ll get pv_pred (per-scene labels), arc_summary (global peak/valley), and score curves.
Assistive only; human-in-the-loop. No model weights or training recipes are exposed.

⚠️ This repository is no longer maintained.
👉 Please visit the new repository: Narrative Engine Emotion (7c)

Latest: v1_0 · vol01 · ep001 — 7c updated (N scenes). No-fee • Research-only • Access request required.

No-fee (access request). Research-only. Includes 5c columns plus conf and plot_break. For a public sample, see the 5c Research Pack.

This is the 7c Core Stable edition (Free, Gated) under the JsisOn License (ARR). It contains the emotion curve only (CSV) with the same 7 fields across releases; schema-stable and reproducible. For open exploration see 5c Free; for richer yet stable coverage, 15c Extended (superset).

Anchor Non-Disclosure (ARR)
This edition does not include numeric lexicon anchors. Per-scene VAD values are pre-snap continuous estimates, not label anchor coordinates.

Why 7c (vs 9c)

Fixed 7 columns, zero-NA policy, range-checked → downstream-stable.
Internal thresholds/distances are not exposed → IP-safe.
Curve-only CSV; no JSONL scenes or full text.
Fixed lexicon version (JS-LEX core100 v1) → stable labels across releases.
Labels use stable identifiers; the mapping is provided in lexicon/core100.v1.bin.tsv.

Schema (7 columns)

Sample (CSV)

scene_idx,valence,arousal,dominance,label,conf,plot_break 1,0.5,0.5,0.5,longing,0.595939,1 2,0.7,0.6,0.6,interest,1,0 3,0.2,0.8,0.4,tension,1,0 4,0.7,0.6,0.6,interest,1,0 5,0.45,0.875,0.35,astonishment,0.589674,0

Fields
- scene_idx (int) — 1-based index within the episode
- label (string) — primary emotion from JS-LEX core100 v1 (see lexicon/core100.v1.bin.tsv)
- conf (float, 0..1) — confidence for label
- valence (float, 0..1)
- arousal (float, 0..1)
- dominance (float, 0..1)
- plot_break (boolean: 0/1)

Files

data/7c_curve.csv — per-scene emotion curve (7 cols)
schema/public_contract.schema.json — public interface (no implementation details)
docs/LICENSE.txt — JsisOn License (ARR) terms
docs/GATED_POLICY.md — access terms
lexicon/core100.v1.bin.tsv

Lexicon (core100.v1)

We provide a minimal, non-numeric VAD lexicon to ensure reproducibility without exposing raw anchors.
File: lexicon/core100.v1.bin.tsv (UTF-8, TSV)
Columns: label, V_bin, A_bin, D_bin, direction
Bins: VL, L, M, H, VH (five ordered levels around the neutral anchor)
Neutral anchor: conceptually (0.5, 0.5, 0.5) — choose midpoints per your protocol.
Version: core100.v1
License: research-only (no redistribution / no training without consent)
Exact numeric anchors are reserved to the Pro edition;
attempts to reconstruct anchors are not permitted under ARR.

Reconstruction (optional, by user)
Researchers may map bins to numeric midpoints (e.g., a 5-level scale) in their own code/protocol to derive approximate VAD values for analysis. This dataset intentionally does not include numeric coordinates.

Dataset Summary

7-column VAD core for research and prototyping. No source text. Derived from Batalstone vol.1 pipeline ([email protected], JS-LEX core100).

Access & Gating

This is a Gated Free dataset under the JsisOn License (ARR).

Allowed (non-commercial only): academic research, evaluation, and model training for research.
Prohibited: commercial training/deployment, resale, sublicensing, redistribution.
Access is free after approval on the Hub (request access and accept ARR terms).

Data Structure

.
├─ data/
│  └─ 7c_curve.csv          
├─ schema/
│  └─ public_contract.schema.json  
├─ lexicon/
│  └─ core100.v1.bin.tsv     
├─ docs/
│  ├─ LICENSE.txt           
│  └─ GATED_POLICY.md

VAD fields are pre-snap continuous estimates;
numeric lexicon anchors are not part of this edition.

B. Schema (public)

schema/public_contract.schema.json

{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "$id": "https://schemas.jsison.com/batalstone/curve_7c_core.arr.v1.json",
  "title": "Batalstone 7c Core Stable — Curve Only (ARR)",
  "description": "No numeric lexicon anchors are included; VAD fields are pre-snap continuous estimates.",
  "type": "object",
  "required": ["scene_idx","valence","arousal","dominance","label","conf","plot_break"],
  "properties": {
    "scene_idx": { "type": "integer", "minimum": 1, "description": "1-based scene index" },
    "valence":   { "type": "number",  "minimum": 0.0, "maximum": 1.0 },
    "arousal":   { "type": "number",  "minimum": 0.0, "maximum": 1.0 },
    "dominance": { "type": "number",  "minimum": 0.0, "maximum": 1.0 },
    "label":     {
      "type": "string",
      "description": "Primary emotion (JS-LEX core100 v1). See lexicon/core100.v1.bin.tsv (non-numeric bins)."
    },
    "conf":      { "type": "number",  "minimum": 0.0, "maximum": 1.0 },
    "plot_break":{
      "description": "0/1 in CSV; cast to boolean by readers",
      "oneOf":[{"type":"integer","enum":[0,1]},{"type":"boolean"}]
    }
  },
  "additionalProperties": false
}

from datasets import load_dataset, Features, Value
features = Features({
  "scene_idx": Value("int32"),
  "valence":   Value("float32"),
  "arousal":   Value("float32"),
  "dominance": Value("float32"),
  "label":     Value("string"),
  "conf":      Value("float32"),
  "plot_break":Value("bool"),
})
ds = load_dataset("jsisonou/webnovel-narrative-emotion",
                  data_files="data/7c_curve.csv",
                  split="train",
                  features=features)

Edition	Columns	Access	Purpose	Why choose it
9c Preview (archived)	9 + diagnostics	archived	historical preview	experimental fields; unstable
7c Core (gated, free)	7	gated-free (ARR)	research & evaluation standard	stable schema, zero-NA, fixed lexicon, reproducible
5c Free	5	public	quick baselines	ultra-light, open

Intended Use & Limitations

Research, benchmarking, prototyping. Non-commercial model training and evaluation are allowed; any commercial use requires a separate written agreement (see License).

License

LicenseRef-JsisOn (ARR) — full text in docs/LICENSE.txt.
Summary: https://js-is-on.com/licenses/1.0/
(Non-commercial research/review only; no redistribution, sublicensing, or commercial training/deployment.)

Contact: [email protected] → 5c Research Pack

Changelog

1.0.0 — Initial 7c Core Stable (no-TH). Supersedes legacy 9c preview.

@dataset{batalstone_emotion_7c_core_arr_2025, author = {Liia Black}, title = {Webnovel Narrative Emotion — 7c Core Stable (Gated, Free)}, year = {2025}, publisher = {JsisOn OÜ} }

Downloads last month: 20