AI4BharatIndicNLG
MultiIndic Paraphrase Generation Unified Script
This paraphrasing release fine-tunes IndicBART on the corresponding IndicNLG dataset. It supports 11 Indian languages.
The task fine-tune is separate from the pretrained model
Its unified-script design follows the base model. Non-Devanagari input and output require script conversion. The release is intended for paraphrasing; it is not a general conversation model.MultiIndic Paraphrase Generation Unified Script official model card
Review the generated text against its input
Practical guidance. Compare meaning as well as wording. Keep negation, quantities and named entities unchanged when a paraphrase must preserve the original claim.
Specifications & access
| Detail | Value and conditions |
|---|---|
| Access | Downloadable |
| Model identifiers | MultiIndicParaphraseGeneration |
| Input scripts | Indian-language text is mapped to Devanagari.MultiIndic Paraphrase Generation Unified Script official model card |
| Training set | 5.53 million sentencesMultiIndic Paraphrase Generation Unified Script official model card |
| License | MIT in the model card.MultiIndic Paraphrase Generation Unified Script official model card |
Related models
Multilingual generation model that maps Indian-language text into shared Devanagari script.
Launch date not confirmedDownloadableSeparate-script IndicBART adaptation for paraphrasing text in 11 Indian languages.
Launch date not confirmedDownloadable