Skip to content

premkumar2306/open-indus-lab

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1,383 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

Open Indus Lab

An open computational framework for testing and comparing Indus-Harappan script decipherment hypotheses.

Live platform: https://premkumar2306.github.io/open-indus-lab/ · Paper (final draft, PDF): download · Citable dataset: v1.0.0 release

Current Reading Layer Status

Note from the author (Ponmuthu Shanmugham): "I haven't completed the English translations as I didn't need them for the research, but they will help non-Tamil speakers. It needs two levels — phoneme (sounds) and morphemes (meaning). It needs to be done, but I haven't."

Layer What Status
Layer 1 — Phoneme Tamil script + Roman transliteration (the sounds) ✅ Complete for all 204 readings shown
Layer 2 — Morpheme English meaning (the translation) 🔲 40 complete (sourced from the final paper, Table 2), 164 pending

Tamil-speaking researchers can evaluate all 204 readings now via Layer 1. Non-Tamil researchers can evaluate the 40 complete readings and the full methodology. See docs/READING_LAYERS.md for full explanation.


⚠️ This platform does not claim to have decoded the Indus script. It provides reproducible, open tools for evaluating competing hypotheses using corpus data, rule engines, statistics, and peer review workflows.

Data provenance (single source of truth)

Every seal reading and sign mapping is sourced directly from the author's documents (INDUS_SEALS_READ_in_book_5, Indus_Signs_Reading, and the draft article). The canonical tables are data/seed/seals_master.csv (194 seal readings) and data/seed/signs_master.csv (264 sign→phoneme mappings). The external Parpola/CISI corpus and the guessed Parpola→Mahadevan sign bridge have been removed — see docs/VALIDATION_REBUILD_REPORT.md.

How to cite

Shanmugham, P. (2026). Open Indus Lab: Indus-Harappan Seal Readings and Sign-Phoneme Mappings (Ancient Tamil Hypothesis), v1.0.0. https://github.com/premkumar2306/open-indus-lab

Stable dataset snapshots are published as GitHub Releases; cite the release version you used. Machine-readable metadata: CITATION.cff.


Project Status

Stage Name Status
Stage 1 Data model & seed schema ✅ Complete
Stage 2 Author-direct seed data (external corpus removed) ✅ Complete
Stage 3 Research dashboard (React, live on GitHub Pages) ✅ Complete
Stage 4 Citable dataset release (v1.0.0, CITATION.cff) ✅ Complete
Stage 5 Final paper ingestion (VER_4 Table 2 → Layer 2) ✅ Complete
Stage 6 Preprint + journal submission (OSF / HSSC / harappa.com) 🔶 In progress
Stage 7 Sign numbering resolved: book numbering (342=aa) is canonical; paper Table 1 to be corrected by author ✅ Complete
Stage 8 Statistical validation & rule engine 🔲 Planned

See FULL_PLAN.md for complete roadmap.


Background

The Indus-Harappan script (~3300–1300 BCE) remains undeciphered after over 100 years of scholarship. More than 4,000 inscribed objects have been found across a 1.3 million sq km civilisation.

This platform provides a neutral computational layer for researchers to:

  • Load and explore the open seal corpus
  • Register named, versioned decipherment hypotheses
  • Apply deterministic rule engines to produce candidate readings
  • Compare hypotheses statistically against null baselines
  • Invite peer reviewers and track objections systematically

Primary Research Input

Author: Ponmuthu Shanmugham (Member of Technical Staff, Retd., Lucent Technologies – Bell Laboratories) Contact: ponmuthushanmugham@gmail.com

Published Works

Title ISBN
Ancient Harappans Speak! After 5000 years 979-8-9949362-3-5 (pb) / 979-8-9940362-4-2 (eBook)
Indus Script is a Language — A living language of 80 million (Bilingual) 979-8-9940362-6-6 (Kavin Publishers, 2026)
Reading Indus-Harappan Script: Research Keys (99 pages) 979-8-9940362-9-7 (2026)

Draft Article

Indus-Harappan Script: A Multi-Disciplinary Analysis of Linguistic Continuity and Urban Topology (final draft, 2026) — PDF on the platform. Submission targets: Humanities and Social Sciences Communications; harappa.com; Academia.edu.


Data (data/seed/)

Canonical author data (single source of truth)

File Rows Description
seals_master.csv 194 Seal readings transcribed verbatim from INDUS_SEALS_READ_in_book_5
signs_master.csv 264 Sign→Tamil phoneme mappings from Indus_Signs_Reading (all corruption fixes documented in-row)
readings_paper_VER4.csv 40 Romanized readings + English meanings from the final paper (Table 2 & figures)
phoneme_mappings_author_full.csv 264 Working phoneme table (superseded by signs_master.csv)
readings_author_full.csv 206 Working readings table (superseded by seals_master.csv)

Research framework data

File Rows Description
hypotheses.csv 6 Competing hypotheses incl. null (FSW2004)
sites.csv 8 Major excavation sites with GPS
motifs.csv 8 Seal motif catalogue
decoding_rules_VPS2024.csv 11 Deterministic rule engine rules
tally_mark_rules.csv 11 Tally 1–12 Tamil phoneme rules
evidence_links.csv 9 Academic citations (paper endnotes)
readings_VPS2024.csv / readings_enriched_VPS2024.csv / readings_body_english.csv 24 / 206 / 15 Article-derived reading detail
phoneme_mappings_appendixA_full.csv / readings_appendixB_tamil.csv 241 / 116 Research Keys appendices

Removed (2026-07): the external Parpola/CISI corpus (seals.csv, signs.csv, sign_sequences.csv) and the guessed Parpola→Mahadevan bridge. Its numbering had zero verified correspondence to the author's data. See docs/VALIDATION_REBUILD_REPORT.md.

Corpus scale (author clarification)

Each listed seal may represent a group of similar seals (1, 5, 20, 40, 60, 130, up to 400). Seals shown read in the paper: 200+, extrapolating to 700+ including represented groups. Sample counts from the paper: cow 300, milk 44, ghee 24, ordinances 128, children's food 160, manure 60.


The VPS2024 Hypothesis

Core Claims

  1. Phonetics: 10 vowels + 18 consonants mapped to Tamil. Most frequent sign = cow face = phoneme aa (ஆ)
  2. Tally marks: Phonetic placeholders using Tamil numeral names (tally-4 = nankuna → enables ney = ghee)
  3. Motifs as street signs: Unicorn = Market Common; Elephant = Elephant Street; No motif = city-wide ordinance
  4. Modifiers: Have form but no independent sound; alter adjacent sign phoneme
  5. Compound integration: Signs 10+12 → sign 15 (aakaavva = cow carer), documented in seal 3246

Key Seal Readings

Seal Motif Reading Meaning
3023, 2358 Unicorn didiir aa Just expressed fresh milk
1076 Unicorn thava thava ney Very good ghee
2082 Unicorn Isa muu aagaavva Cow carer of 3 cows of Isa
2127 Elephant eNNey Sesame oil
2234 Unicorn+Bull vayya yiir illam The world is a Big Home
4718 None ay ya ay o vyamun; ammu Panic milk for children (ordinance)
1425 None illam vaya muu thani Homes for the aged and alone
2444 Unicorn aNNal kai aaNai Order of the Leader
Dholavira None Residence of king's advisors + grain store

Competing Hypotheses

Code Researcher Claim Role
VPS2024 Ponmuthu Shanmugham Ancient Tamil, syllabic Primary input
JEEVA2020 Purnachandra Jeeva Ancient Tamil (prior work) Extended by VPS2024
PARPOLA Asko Parpola Proto-Dravidian, logo-syllabic Comparison
MAHADEVAN Iravatham Mahadevan Dravidian concordance Comparison
FSW2004 Farmer, Sproat, Witzel Non-linguistic emblems Null hypothesis
RAO2009 Rajesh Rao et al. Language-like entropy Statistical baseline

Tech Stack

Layer Technology
Frontend React 18 + Vite, deployed to GitHub Pages (legacy mode, served from main:/docs)
Data Versioned CSVs in data/seed/ (canonical), validated every 4 hours by GitHub Actions
Backend (scaffold) Python rule engine + PostgreSQL schema for the planned statistical stage
Validation scripts/validate.pydocs/VALIDATION_REPORT.md (currently 16 pass / 0 fail)

Corpus Sources

Source Content License
mayig/indus-valley-script-corpus Removed 2026-07 — incompatible numbering; see SOURCES.md MIT
RMRL Chennai — indusscript.in Seal images, Mahadevan concordance Research use
CISI Vol.1 — Joshi & Parpola, Helsinki 1987 Collections in India Reference
CISI Vol.2 — Shah & Parpola, Helsinki 1991 Collections in Pakistan Reference
Dr. N. Yadav et al. — Harappa.com Sign frequency data Reference
Rajan & Sivamantham, Govt of Tamil Nadu (2025) Graffiti mark morphology Reference

License

MIT. Corpus data credit to respective sources above.

Research content © Ponmuthu Shanmugham 2026. Used with permission for open research and peer review.