Indic languages are written in a variety of scripts (Brahmi family of abugida scripts, Arabic-derived abjad scripts, and even alphabetic Roman script). This diversity makes it challenging to support mechanisms which are convenient for typing or creating content in these diverse languages and scripts. Most Indian users are comfortable with the Roman keyboard and thus an optimal solution that users find beneficial is automatic transliteration of the romanized input into the native script. To enable this, at AI4Bharat, we have undertaken the task of creating large-scale transliteration corpora for Indic languages along with models for transliteration of romanized inputs into native scripts.
Bhasha-Abhijnaanam is a language identification test set for native-script as well as Romanized text spanning 22 Indic languages.
BPCC is a comprehensive and publicly available parallel corpus containing a mix of Human labelled data and automatically mined data; totaling to approximately 230 million bitext pairs.
A multilingual transformer based model for transliteration from romanized input to native language scripts supporting 21 languages. This model is trained using Aksharantar corpus and at the time of its release was the state of the art open source model as evaluated on Google's Dakshina benchmark and our Aksharantar benchmark.
A model for identifying the language of romanized content on social media. This model will be trained using the large number of romanized Indian language words in Aksharantar.
A transliteration interface (En-Indic online keyboard) that converts romanized text into Indic text as you type.
A transliteration interface that converts Indic text into romanized text and vice versa.
An on-going research for creating swipe-based keypad typing on Android devices for Indic languages
We have partnered with Pratham Books to enable romanized keyboards in their translation interface for low resource languages such as Bodo, Kashmiri, Konkani, Maithili, Nepali and Urdu.