Hugging Face's logo

Datasets:
lsadouk1111
/
QESC

Tasks:
Text Retrieval
Text Classification
Modalities:
Tabular
Text
Formats:
json
Languages:
English
French
Arabic
Size:
< 1K
Tags:
Islamic education
emotional support
children
Arabic
Moroccan Darija
NLP
Libraries:
Datasets
pandas
Polars
License:
Dataset card Data Studio Files
xet
 Community
QESC / README.md
lsadouk1111's picture
lsadouk1111
Update README.md
0322dd0 verified
preview code
|
Raw
History Blame Contribute Delete
7.53 kB
metadata 
language:
  - en
  - fr
  - ar
license: cc-by-nc-4.0
task_categories:
  - text-retrieval
  - text-classification
pretty_name: QESC - Quranic Emotional Situation Corpus
size_categories:
  - n<1K
tags:
  - Islamic education
  - emotional support
  - children
  - Arabic
  - Moroccan Darija
  - NLP
  - retrieval
  - CASS
  - prophet stories
  - emotional intensity

QESC — Quranic Emotional Situation Corpus

Dataset Description

QESC is the first structured corpus of Quranic prophet situational scenes annotated for emotional intensity and narrative resolution type, designed for child-facing emotional support retrieval in multilingual contexts (English, French, Moroccan Darija).

The corpus supports the CASS (Context-Aware Semantic Similarity) framework — a lightweight safety-aware retrieval system that matches a child's free-text emotional expression to the most contextually appropriate prophet story, while preventing emotionally dangerous mismatches.

  • Paper: CASS: A Context-Aware Semantic Similarity Framework for Safe Retrieval in Child-Facing Educational Applications (Sadouk and Gadi)
  • GitHub: https://github.com/lsadouk/CASS
  • Application: Ya Allah — Islamic children's emotional support app

Dataset Summary

Property Value
Total entries 100
Quranic figures covered 26
EI scale 1–5 (mild → severe)
RT scale 1–3 (immediate → delayed)
Primary language English
Additional languages Arabic (names), French, Moroccan Darija (paraphrases)
License CC BY-NC 4.0

Supported Tasks

  • Emotional situation retrieval — match a child's free-text emotional query to the most appropriate prophet story
  • Emotional intensity classification — predict EI level from situation description
  • Resolution type classification — predict RT from situation description

Dataset Structure

Fields

Field Type Description
id string Unique identifier e.g. moses_003
prophet_latin string Prophet name in English e.g. Moses
prophet_arabic string Prophet name in Arabic e.g. موسى
situation_en string 3–5 sentence English description of the emotional situation
emotions list[string] 2–5 primary emotion labels
EI float Emotional Intensity 1.0–5.0 (half-point increments)
RT int Resolution Type: 1=immediate, 2=gradual, 3=delayed
RT_label string Human-readable RT label
child_paraphrase string Child-friendly multilingual explanation and lesson
quran_ref string Quranic verse or hadith reference
RL int Recitation Level 1–3
gender_of_protagonist string male or female
annotator_agreed bool Annotation confirmed by domain reviewer
version string Corpus version

EI Distribution

EI Label Count
1 Mild everyday emotion 19
2 Low-moderate distress 19
3 Moderate distress 19
4 High distress 24
5 Severe crisis-level distress 19

RT Distribution

RT Label Count
1 Immediate resolution 27
2 Gradual resolution 29
3 Delayed resolution (long patient endurance) 23

Quranic Figures Covered

Prophets: Adam, Idris, Noah, Hud, Salih, Abraham, Lot, Ishmael, Isaac, Jacob, Joseph, Job, Shuaib, Moses, Aaron, David, Solomon, Elijah, Elisha, Jonah, Zachariah, John, Jesus, Muhammad.

Key figures (non-prophets): Mary, Hagar.

Example Entry 

{
  "id": "joseph_003",
  "prophet_latin": "Joseph",
  "prophet_arabic": "يوسف",
  "situation_en": "Joseph was separated from his beloved father Jacob
    and thrown into a well by his own brothers out of jealousy.
    He was then sold into slavery in Egypt, far from everything
    he loved. Despite years of separation, Joseph never lost hope
    in Allah or stopped praying for his family.",
  "emotions": ["grief", "loneliness", "abandonment", "yearning", "patient_hope"],
  "EI": 3.0,
  "RT": 3,
  "RT_label": "delayed_resolution",
  "child_paraphrase": "Prophet Yusuf missed his father every single
    day for many years. He never stopped loving him even when they
    were far apart. Allah never forgot Yusuf, and one day reunited
    them in a way more beautiful than either could have imagined.",
  "quran_ref": "12:15-18",
  "RL": 2,
  "gender_of_protagonist": "male",
  "annotator_agreed": true,
  "version": "1.0"
}

Usage 

from datasets import load_dataset

ds = load_dataset("lamyaa/QESC")

# All entries
print(f"Total entries: {len(ds['train'])}")

# Filter by EI level
mild = [e for e in ds['train'] if e['EI'] <= 2.0]
severe = [e for e in ds['train'] if e['EI'] == 5.0]

# Filter by prophet
joseph = [e for e in ds['train'] if e['prophet_latin'] == 'Joseph']

# Filter by resolution type
immediate = [e for e in ds['train'] if e['RT'] == 1]

Use with CASS retrieval 

# See full code in: https://github.com/lsadouk/CASS

# English query
retrieve("I feel sad and nobody understands me")
# → joseph_003 (EI=3, RT=3)

# French query
retrieve("Je me sens abandonné par mes amis")
# → muhammad_002 (EI=3, RT=2)

# Moroccan Darija query
retrieve("ma kaynch had li yfahmni")
# → shuaib_001 (EI=3, RT=3)

Safety Properties

QESC is designed with child safety in mind. The EI annotation enables the CASS framework to prevent emotionally dangerous matches:

  • A child at EI=1 (mild embarrassment) will never receive an EI=5 story (extreme suffering)
  • A child at EI=5 (severe crisis) will never receive an EI=1 story (gentle everyday lesson)
  • Safety filtering is automatic — no user configuration required

Across 5 diagnostic queries, CASS suppressed 27 of 28 emotionally dangerous entries with a mean rank drop of 37 positions and maximum of 62 positions.

Annotation Guidelines

EI scale:

  • EI=1: Mild everyday emotion — embarrassment, small disappointment, gentle worry
  • EI=2: Low-moderate distress — social exclusion, mild loneliness, minor conflict
  • EI=3: Moderate distress — betrayal, sustained loneliness, public humiliation
  • EI=4: High distress — acute fear, grief, unjust accusation, family conflict
  • EI=5: Severe crisis — extreme suffering, hopelessness, years of loss

RT scale:

  • RT=1: Immediate resolution — comfort or answer arrives quickly (within the same scene)
  • RT=2: Gradual resolution — situation improves over weeks or months
  • RT=3: Delayed resolution — resolution requires long patient endurance (years)

Citation 

@article{sadouk2025cass,
  title     = {CASS: A Context-Aware Semantic Similarity Framework
               for Safe Retrieval in Child-Facing Educational Applications},
  author    = {Sadouk, Lamyaa AND Gadi, Taoufiq},
  journal   = {Education and Information Technologies},
  year      = {2025},
  publisher = {Springer}
}

Dataset Card Authors

Lamyaa Sadouk Ecole Marocaine des Sciences de l'Ingénieur, Casablanca, Morocco

Taoufiq Gadi Laboratoire de Recherche en Mathématique, Informatique et Sciences de l'Ingénieur (MISI), Morocco

License

CC BY-NC 4.0 — Attribution-NonCommercial 4.0 International

You may use and share this dataset for research and educational purposes with attribution. Commercial use is not permitted without explicit permission from the author.