项目作者: sagorbrur

项目描述 :
transliteration for indic language
高级语言: Python
项目地址: git://github.com/sagorbrur/itranslit.git
创建时间: 2021-04-29T11:48:13Z
项目社区:https://github.com/sagorbrur/itranslit

开源协议:MIT License

下载


iTRANSLIT

iTRANSLIT is a deep learning based transliteration package for indic language

Installation

pip install itranslit

Dependency

  • pytorch 1.7.0 or 1.7.0+

NB: No GPU need. It’s CPU based.

Supported Language and Language Code

Language Name Langauage Code
Bangla bn
Gujarati gu
Hindi hi
Punjabi pa
Sindhi sd
Urdu ur
Malayalam ml
Tamil ta

API

  1. from itranslit import Translit
  2. translit = Translit('bn')
  3. word = "aami"
  4. output = translit.predict(word, topk=10)
  5. print(output)

Datasets and Training Details

  • We used Google Dakshina Dataset
  • Thanks to AI4Bharat for providing training notebook with details explanation
  • We trained Google Dakshina lexicons train datasets for 10 epochs with batch size 128, 1e-3, embedding dim = 300, hidden dim = 512, lstm, used attention
  • We evaluated our trained model with Google Dakshina lexicon test data using AI4Bharat evaluation script
  • You can find evaluation summary here