IndicXlit
A Transformer-based multilingual transliteration model
About Model
Bhashini - IndicXlit is a Transformer-based multilingual transliteration model, trained on Aksharantar dataset which is the largest publicly available parallel transliteration corpora collection for Indic languages at the time of writing (20 May 2022). It is used to convert any roman text written in Indian language (like Hinglish) to the native Indic-script (like Devanagari for Hindi). It supports 21 Indic languages: Assamese, Bangla, Bodo, Gujarati, Hindi, Kannada, Kashmiri, Konkani, Maithili, Malayalam, Manipuri, Marathi, Nepali, Oriya, Panjabi, Sanskrit, Sindhi, Sinhala, Tamil, Telugu, Urdu.
IndicXlit
Metadata
MIT
AI4Bharat
Machine Translation Model
Other
Open
Sector Agnostic
05/03/25 15:25:05
3.94 MB
IndicXlit-master ( 3 files, 12 directories )
ablation_study
2 directories
app
10 files, 1 directories
Checker
3 files
corpus_preprocessing
5 directories
data_mining
1 files, 2 directories
Dataset_Format
2 files
inference
2 directories
model_training_scripts
1 files, 7 directories
.gitignore
1.79 KB
LICENSE
1.04 KB
This preview shows 10 out of 15 items. Load more
Activity Overview
0
28 
3.94 MB 
907
Tags
- Language Modeling
- Multilingual Translation
- Machine Translation
- Regional Languages
- Indian Languages
- NLP
- transliteration
License Control
MIT
Version Control
Version 1(3.94 MB)
- admin·1 year(s) ago
- chevron_right
IndicXlit-master
- chevron_right
ablation_study
- chevron_right
app
- chevron_right
Checker
- chevron_right
corpus_preprocessing
- chevron_right
data_mining
- chevron_right
Dataset_Format
- chevron_right
inference
- chevron_right
model_training_scripts
.gitignore
LICENSE
- more_horiz 5 more
More Models from TechCorp
- 0
- 17
- 115.62 MB
102
- 0
- 13
- 2.08 GB
- 164
- 0
- 7
- 2.09 GB
- 106
- 0
- 3
- 3.52 GB
- 104
- 0
- 3
- 3.52 GB
- 95
- 0
- 4
- 3.52 GB
- 93
- 0
- 5
- 3.52 GB
- 134
- 0
- 3
- 3.52 GB
- 155
- 0
- 7
- 3.52 GB
- 152
- 0
- 5
- 3.52 GB
- 154
© 2026 - Copyright AIKosh. All rights reserved.