India's 22 languages connected by a massive 1 million sentence bridge
bharatSetu: a 1M x 22 bridge for indic translation
Computation and LanguageArtificial Intelligence
Summary
Many Indian languages lack direct translation examples between each other, which makes it hard for computers to translate. The authors created BharatSetu, a large table linking nearly a million English sentences to machine translations in all 22 officially recognized Indian languages. This means any two languages can be connected by translating through English, even if no direct examples exist between them. Because the translations are machine-generated and not checked by humans, the authors suggest ways to track confidence and fix errors over time. They did not test or report translation quality yet.
What this means in practice
- •For machine translation engineers: Use a large aligned dataset spanning 22 Indian languages to train or improve translation systems covering low-resource pairs without direct bilingual data.
- •For natural language processing developers: Integrate a machine-generated multilingual resource for Indian languages to create new cross-lingual applications or services in multilingual contexts.
Authors
Haq Nawaz Malik, Nahfid Nissar, Faizan Iqbal
Abstract
India's Eighth Schedule recognises 22 languages, yet parallel data is uneven across them, and many language pairs, especially among lower-resource languages such as Bodo, Dogri, Santali and Kashmiri, lack direct bitext. We present BharatSetu 22-Indic 1M, a row-aligned table of 984,764 English sentences, each paired with machine translations into all 22 scheduled languages (21,664,808 Indic cells), generated with the sarvamai/sarvam-translate model family. Because every Indic cell in a row translates the same English sentence, any two columns form a translation pair: 462 directed Indic-to-Indic pairs, or 506 including English, without direct bitext. We call these pivot-derived pairs and propose a confidence layer that lets users choose how strictly each pair is connected. The corpus is machine-generated and not human-verified; quality is expected to vary by language, domain and sentence style, and pivot-derived pairs can inherit errors from both sides. We therefore release BharatSetu as a machine layer and propose a human-correction layer keyed to the same row identifiers, so that each corrected column upgrades every pair that uses it, together with an evaluation protocol. We report no translation-quality results.