📚 BCS বাতিঘর — Knowledge Graph Pipeline

পথ দেখায় প্রস্তুতির • Interactive Pipeline Dashboard
🧠 Souvik — KG Schema & Construction 💾 Saif — Episodic Memory Layer ✅ Galib — Fact Quality Gate

📊 Pipeline Overview

📊 Node Type Distribution

🔗 Edge Type Distribution

✅ Quality Gate Results

📝 MCQ Suitability Breakdown

🕸️ Interactive Knowledge Graph

Entity Fact Source Topic Question

Hover for details. Scroll to zoom. Drag to pan. Click to focus.

📋 Fact Explorer

#TopicFact TextScoreVerdictMCQMCQ Types

✅ Fact Quality Gate — Galib

📈 Score Distribution

🎯 Quality Dimensions (Average)

📝 Quality Scoring Weights

📊 Topic Density Analysis

📊 Facts & Questions per Topic

🎯 Topic Performance (Episodic)

TopicFactsQuestionsUnmapped FUnmapped QRecommendation

🧠 Episodic Memory — Saif

Episodes
Failed Facts
Memory Lifecycle
Episode IDTopicScoreStatusDecay
Fact IDRejectionsPreview

Memory Lifecycle

1. WriteEpisodes created when MCQs generated from KG facts 2. RetrieveSimilar episodes queried by topic + score threshold 3. UpdateJudge feedback updates acceptance status & rejection logs 4. ForgetDecay formula: e^(-0.05 * age_days) with threshold-based pruning 5. PreserveEpisodes with accepted=1 AND score>0.85 are never pruned

👥 Entity Explorer

#NameSubtypeNode ID

⚙️ Pipeline Architecture

KG Schema — Node Types

ENTITY Real-world named entities (PERSON, COUNTRY, CITY ...)
FACT Atomic factual claims with quality scores
SOURCE URLs and publishers
TOPIC Knowledge domains (History, Geography ...)
QUESTION Generated or refined questions

Edge Types (Relations)

SUBJECT_OF Entity to Fact
OBJECT_IS Fact to Entity
SUPPORTED_BY Fact to Source
ABOUT Fact/Question to Topic
ASKS_FOR Question to Fact
DERIVED_FROM Question to Question

File Inventory

Source Code
kg_builder.py — KG construction (Souvik)
fact_quality.py — Quality gate (Galib)
episodic_store.py — Memory layer (Saif)
sample_fact_generator.py — Fact data
pipeline.py — Full pipeline script
generate_dashboard.py — Dashboard generator
Documentation
kg_schema.md — KG schema spec
memory_lifecycle.md — Memory design
quality_report.txt — Quality analysis
Artifacts
bcg_gk_facts.json — Fact dataset
memory.db — Episodic memory DB
snapshots/*.gpickle — KG snapshots
pipeline.ipynb — Notebook

Entity Subtypes (18)

PERSONORGANIZATION COUNTRYPLACE CITYRIVER MOUNTAINFOREST OCEANINSTITUTION EVENTDOCUMENT TREATYAWARD LANGUAGECURRENCY ANIMALPLANT

📂 Raw Data & Export

📥
Download Facts JSON
📥
Download Entities JSON
📥
Download Episodes JSON
📥
Download All Data

Quick Stats