Bhashini-AI4Bharat Textual Language Detection v1.0

About Model

IndicLID, is a language identifier for all 22 Indian languages listed in the Indian constitution in both native-script and romanized text. IndicLID is the first LID for romanized text in Indian languages. It is a two stage classifier that is ensemble of a fast linear classifier and a slower classifier finetuned from a pre-trained LM. It can predict 47 classes (24 native-script classes and 21 roman-script classes plus English and Others). IndicLID is evaluated on Bhasha-Abhijnaanam benchmark which is released alnog with this work. For native-script text, IndicLID has better language coverage than existing LIDs and is competitive or better than other LIDs. IndicLID model is 10 times faster and 4 times smaller than the NLLB model also establish a strong baseline results on the roman-script text.