My Model Thought Tamil Was Telugu: Open-Source Language ID for Hinglish, Tanglish and Kanglish
Last Updated on November 2, 2026 by Editorial Team
Author(s): Latent AI Labs
Originally published on Towards AI.
My Model Thought Tamil Was Telugu: Open-Source Language ID for Hinglish, Tanglish and Kanglish
By Arun Prabhakaran · @LatentAILabs

The article introduces Bhasha, an open-source word-level language identification model for code-mixed Indian text (including Hinglish and other romanized/native-script mixes). It explains why message-level language ID is insufficient when you need per-word language tags for tasks like transliteration, tokenization, translation modeling, measuring code-mixing, and filtering corpora. The author describes how Bhasha was built from 1.65M code-mixed sentences generated programmatically from public text, and how the model outputs both word-level labels and message-level summaries (main language, script type, and language shares). Results from various evaluation sets show strong performance overall, but a surprising weakness for Tamil: on human-labelled data, many Tamil words are incorrectly tagged as Telugu, a failure attributed to differences between written/news Tamil (which dominated training) and spoken forms common in comments. A “language-pair hint” (the langs option) helps, and Bhasha v2 improves Tamil substantially by adding more real, word-labelled YouTube comment data. The post closes with guidance on where Bhasha works best, key lessons from the Tamil error analysis, and release details plus links to try Bhasha and related models.
Read the full blog for free on Medium.
Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.
Published via Towards AI
Towards AI Academy
We Build Enterprise-Grade AI. We'll Teach You to Master It Too.
15 engineers. 100,000+ students. Towards AI Academy teaches what actually survives production.
Start free — no commitment:
→ AI Engineering For Production — our new book, coming before the end of 2026. Join the waitlist.
→ Agents Architecture Cheatsheet — 3 years of architecture decisions in 6 pages
Our courses:
→ AI Engineering Certification — 90+ lessons from project selection to deployed product. The most comprehensive practical LLM course out there.
→ Agent Engineering Course — Hands on with production agent architectures, memory, routing, and eval frameworks — built from real enterprise engagements.
→ AI for Work — Understand, evaluate, and apply AI for complex work tasks.
Note: Article content contains the views of the contributing authors and not Towards AI.