Malayalam (മലയാളം) Classifier using fastai (Working in Progress)
🥳 This model is my attempt to use machine learning using Malayalam Language. Huge inspiration from Malayalam Text Classifier. Courtesy to @waydegilliam for blurr
🌈 മലയാളത്തിൽ മെഷീൻ ലീർണിങ് പഠിക്കാനും പിന്നേ പരിചയപ്പെടാനും, to be continued...
How its built ? & How to use ?
Please find the notebook used for training the model
Usage:
First, install the utilities to load the model as well as blurr
, which was used to train this model.
!pip install huggingface_hub[fastai]
!git clone https://github.com/ohmeow/blurr.git && cd blurr && pip install -e ".[dev]"
from huggingface_hub import from_pretrained_fastai
learner = from_pretrained_fastai("rajeshradhakrishnan/ml-news-classify-fastai")
sentences = ["ഓഹരി വിപണി തകരുമ്പോള് നിക്ഷേപം എങ്ങനെ സുരക്ഷിതമാക്കാം",
"വാര്ണറുടെ ഒറ്റക്കയ്യന് ക്യാച്ചില് അമ്പരന്ന് ക്രിക്കറ്റ് ലോകം"]
probs = learner.predict(sentences)
# 'business', 'entertainment', 'sports', 'technology'
for idx in range(len(sentences)):
print(f"Probability that sentence '{sentences[idx]}' is business is: {100*probs[idx]['probs'][0]:.2f}%")
print(f"Probability that sentence '{sentences[idx]}' is entertainment is: {100*probs[idx]['probs'][1]:.2f}%")
print(f"Probability that sentence '{sentences[idx]}' is sports is: {100*probs[idx]['probs'][2]:.2f}%")
print(f"Probability that sentence '{sentences[idx]}' is technology is: {100*probs[idx]['probs'][3]:.2f}%")
Model card
Model description
The is a Malayalam classifier model for labels 'business', 'entertainment', 'sports', 'technology'.
Intended uses & limitations
The model can be used to categorize malayalam new sfeed.
Training and evaluation data
Data is from the AI4Bharat-IndicNLP Dataset and wrapper to extract only Malayalam data( HF dataset)!.
Citation
@article{kunchukuttan2020indicnlpcorpus,
title={AI4Bharat-IndicNLP Corpus: Monolingual Corpora and Word Embeddings for Indic Languages},
author={Anoop Kunchukuttan and Divyanshu Kakwani and Satish Golla and Gokul N.C. and Avik Bhattacharyya and Mitesh M. Khapra and Pratyush Kumar},
year={2020},
journal={arXiv preprint arXiv:2005.00085},
}