FAQ

Basics

Has the Indus script been deciphered?: No. There is no universally accepted decipherment. This site will not claim otherwise.
What language did the Indus civilisation speak?: Unknown. Several families have been proposed. None is demonstrated.
Can I translate a seal here?: No. The Reading Lab offers structural paraphrase and optional labelled hypotheses, not translations.
What does a fish-shaped sign mean?: Its shape is a visual description only. A Proto-Dravidian mīn reading is a semantic candidate, not a fact.
Are the records here the full historical corpus?: No. Public browse currently uses a demonstration dataset, clearly labelled, until verified imports replace it.
Who is this for?: Curious visitors, students, independent researchers, and — later — institutional partners who need provenance and APIs.

What is an evidence level?: A six-way label from observed through structural, functional, semantic candidate, unverified, to rejected.
What is a null model?: A randomised control. Our default shuffles sign order inside each inscription while keeping length and inventory.
Why reduce duplicates?: A seal and its clay impressions can repeat a sequence without adding an independent composition.
Does high PMI mean a word?: No. PMI flags attraction between signs. It still needs controls and does not yield a gloss.
Why not call a sign a suffix?: End-bias is positional. Linguistic morphology would be a further claim.

Which direction should I read?: Stamp seals are usually treated as right-to-left. Toggle visual versus analytical order in the explorer.
What is a CISI or Mahadevan ID here?: Demo identifier fields. They are not a dump of those published corpora.
Why are images placeholders?: Rights must be clear. Unclear images are not shown; abstract glyphs stand in.
Do motifs translate the text?: No. A unicorn image is not a reading of the accompanying signs.
Why is Mohenjo-daro so frequent?: Excavation and publication intensity, and in this demo a seeded site weight. Raw counts are biased.

Is the script Proto-Dravidian?: That is a leading hypothesis, not a result.
Could it be Indo-Aryan?: That model is listed and treated as under investigation, not as a conclusion.
Could the signs be non-linguistic?: Yes, as a global or partial model. We currently treat a pure emblem reading as a weaker fit to rigid templates, still not disproven for every class.
Does AI change this?: AI can help search and cluster. It cannot invent a bilingual or an agreed lexicon.

What is free?: Public corpus browse, research cases, method pages and basic search. Essential disclaimers are never paywalled.
What does Researcher include?: Saved collections, exports of permitted demo tables, analysis runs and comparison tools.
How do I sign in?: Google, X, or email and password.

Can I submit a decipherment?: You can submit evidence against a specific record. Unsourced “I decoded it” claims are rejected.
How are submissions reviewed?: Contributors submit; reviewers comment; editors publish. Core records are versioned.

May I reuse the demo sequences?: As illustrative demo data with attribution and the non-corpus disclaimer. Not as a museum catalogue.
How do I request takedown?: Use the contact page. Rights-unknown images are not displayed publicly.

What does the Reading Lab output?: A conservative functional paraphrase plus a “why this is uncertain” panel.
Are analysis jobs exact?: They are exact on the loaded dataset. The dataset is currently a demonstration seed.
What is Research Case 001?: An illustrative 033–705 directional-compound workflow, not a historical result.