🏛️ Data Architecture — dhatupatha

This document describes the directory structure, the data model, and the processing pipeline of the dhatupatha edition.

---

1. Project Directory Layout

dhatupatha/
├── assets/
│   └── birchville_logo.png           # Birchville Coat of Arms (Brand Asset)
│
├── data/
│   ├── dhatupatha_canonical.json     # Canonical Pāṇinīya reference database (1.7 MB, 2,259 roots)
│   ├── dhatupatha_complete.json      # Aligner master catalog with page concordance (2.1 MB)
│   ├── fonts/                        # Vendored OpenType fonts (Noto + Libertine)
│   │   ├── NotoSerifDevanagari-Regular.ttf
│   │   ├── NotoSerifDevanagari-Bold.ttf
│   │   ├── LinLibertine_R.otf
│   │   ├── LinLibertine_RB.otf
│   │   └── LinLibertine_RI.otf
│   └── pdf/
│       ├── Dhatu-patha.typeset_edition.pdf # Typographic vector re-edition (2.96 MB, 229 pages)
│       └── Dhatu-patha.typeset.typ         # Typst source code of the edition (1.8 MB)
│
├── docs/                             # Bilingual MkDocs documentation
│   ├── de/                           # German documentation (Default)
│   ├── en/                           # English documentation
│   ├── assets/                       # Logos and brand assets
│   └── stylesheets/                  # Birchville Scholarly Synthesis theme
│
├── scripts/                          # Automation & build scripts
│   ├── build_dhatupatha_groundtruth.py # Generates canonical ground-truth
│   ├── align_dhatupatha.py             # Bilingual facing-page alignment engine
│   └── build_typeset_edition.py        # Typst vector compiler
│
├── viewer.html                       # Standalone QA Viewer (Payer standard)
├── mkdocs.yml                        # Documentation site configuration
├── pyproject.toml                    # Python project configuration
├── requirements.txt                  # Python dependencies
├── LICENSE                           # MIT License
└── README.md                         # Project overview

---

2. The Four-Layer Data Model

Layer 1 — Facsimile Image Base
  Source:   410 historical book pages
  Purpose:  Visual background layer in 1:1 Sandwich PDF and primary citation reference

Layer 2 — Bilingual OCR & Facing-Page Geometry
  Content:  Parallel text layers of odd (Devanāgarī) and even pages (IAST)
  Purpose:  Cross-validation of degraded lead typography and complex ligatures

Layer 3 — Canonical Master JSON & Ground-Truth
  Storage:  data/dhatupatha_canonical.json & data/dhatupatha_complete.json
  Content:  2,259 structured root objects across all 10 Gaṇas with page concordance
  Status:   100.0% complete (0 duplicates, 0 omissions)

Layer 4 — Target Artifacts & Typographic Re-Edition
  Storage:  data/pdf/Dhatu-patha.typeset_edition.pdf & viewer.html
  Content:  Print-ready 229-page vector PDF in Typst and standalone QA Viewer

---

3. End-to-End Processing Pipeline

[Historical Scans (410 pages)]
       │
       ▼
[Step 1: Canonical Ground-Truth] ──► scripts/build_dhatupatha_groundtruth.py
       │                                └── data/dhatupatha_canonical.json (2,259 roots)
       ▼
[Step 2: Bilingual Alignment] ─────► scripts/align_dhatupatha.py
       │                                ├── Parallel state machine (Odd Deva / Even IAST)
       │                                └── Page concordance indexing
       ▼
[Step 3: Master Consolidation] ────► data/dhatupatha_complete.json
       │                                (Single Source of Truth, 2,259 entries)
       ▼
[Step 4: Multi-Artifact Build]
       ├── Typst Vector PDF:     data/pdf/Dhatu-patha.typeset_edition.pdf (~2.4s build)
       ├── Typst Source:         data/pdf/Dhatu-patha.typeset.typ
       └── QA Viewer:            viewer.html (Payer standard)

---

4. Dataset Metrics

Metric	Value	Comment
Scanned Book Pages	410 pages	pp. 1–25 (Front Matter), pp. 26–285 (Main Corpus), pp. 286–410 (Back Matter)
Total Root Count	2,259 roots	Completely captured across all 10 Gaṇas
Gaṇa Classes	10 classes	1. Bhvādi to 10. Curādi
Duplicates / Missing	0 / 0	100.0% aligned against the grammatical canon
Typst Book Length	229 pages	Section-specific layout (1-column prose / 2-column dictionary)
Typst Compile Time	~2.4 seconds	Vector rendering via Typst 0.11+
Font Vendoring	5 OpenType files	Hermetically bundled in data/fonts/ (Zero Tofu)

---

5. Data Model Specification (dhatupatha_complete.json)

Each of the 2,259 entries has the following schema:

{
  "id": 1,
  "gana": 1,
  "gana_name": "भ्वादिगणः",
  "gana_iast": "Bhvādigaṇaḥ",
  "dhatu": "भू",
  "iast_dhatu": "bhū",
  "aupadeshik": "भू",
  "iast_aupadeshik": "bhū",
  "pada": "parasmaipada",
  "settva": "seṭ",
  "karma": "akarmaka",
  "artha": "सत्तायाम्",
  "iast_artha": "sattāyām",
  "artha_english": "to be, to exist",
  "orig_page": 26,
  "commentary_note": ""
}

Field	Type	Description
id	Integer	Unique sequential ID (1 to 2,259)
gana	Integer	Root class number (1 to 10)
gana_name	String	Gaṇa name in Devanāgarī (e.g., भ्वादिगणः)
gana_iast	String	Gaṇa name in IAST (e.g., Bhvādigaṇaḥ)
dhatu	String	Canonical root in Devanāgarī (e.g., भू)
iast_dhatu	String	Root in scientific IAST transliteration
aupadeshik	String	Traditional Aupadeśika citation form with anubandha markers
iast_aupadeshik	String	Aupadeśika form in IAST
pada	String	Voice classification (parasmaipada, ātmanepada, ubhayapada)
settva	String	Augment marker (seṭ, aniṭ, veṭ)
karma	String	Transitivity (sakarmaka, akarmaka, dvikarmaka)
artha	String	Traditional Sanskrit locative meaning (e.g., सत्तायाम्)
iast_artha	String	Locative meaning in IAST (e.g., sattāyām)
artha_english	String	English gloss of basic meaning
orig_page	Integer	Historical scan page number (e.g., 26)
commentary_note	String	Additional grammatical notes