Edit model card

MultiIndicQuestionGenerationUnified

MultiIndicQuestionGenerationUnified is a multilingual, sequence-to-sequence pre-trained model, a IndicBART checkpoint fine-tuned on the 11 languages of IndicQuestionGeneration dataset. For fine-tuning details, see the paper. You can use MultiIndicQuestionGenerationUnified to build question generation applications for Indian languages by fine-tuning the model with supervised training data for the question generation task. Some salient features of the MultiIndicQuestionGenerationUnified are:

  • Supported languages: Assamese, Bengali, Gujarati, Hindi, Marathi, Oriya, Punjabi, Kannada, Malayalam, Tamil, and Telugu. Not all of these languages are supported by mBART50 and mT5.
  • The model is much smaller than the mBART and mT5(-base) models, so less computationally expensive for fine-tuning and decoding.
  • Fine-tuned on large Indic language corpora (770 K examples).
  • All languages have been represented in Devanagari script to encourage transfer learning among the related languages.

You can read more about MultiIndicQuestionGenerationUnified in this paper.

Using this model in transformers

from transformers import MBartForConditionalGeneration, AutoModelForSeq2SeqLM
from transformers import AlbertTokenizer, AutoTokenizer

tokenizer = AutoTokenizer.from_pretrained("ai4bharat/MultiIndicQuestionGenerationUnified", do_lower_case=False, use_fast=False, keep_accents=True)
# Or use tokenizer = AlbertTokenizer.from_pretrained("ai4bharat/MultiIndicQuestionGenerationUnified", do_lower_case=False, use_fast=False, keep_accents=True)

model = AutoModelForSeq2SeqLM.from_pretrained("ai4bharat/MultiIndicQuestionGenerationUnified")
# Or use model = MBartForConditionalGeneration.from_pretrained("ai4bharat/MultiIndicQuestionGenerationUnified")

# Some initial mapping
bos_id = tokenizer._convert_token_to_id_with_added_voc("<s>")
eos_id = tokenizer._convert_token_to_id_with_added_voc("</s>")
pad_id = tokenizer._convert_token_to_id_with_added_voc("<pad>")
# To get lang_id use any of ['<2as>', '<2bn>', '<2gu>', '<2hi>', '<2kn>', '<2ml>', '<2mr>', '<2or>', '<2pa>', '<2ta>', '<2te>']

# First tokenize the input and outputs. The format below is how IndicBART was trained so the input should be "Sentence </s> <2xx>" where xx is the language code. Similarly, the output should be "<2yy> Sentence </s>". 
inp = tokenizer("7 फरवरी, 2016 [SEP] खेल 7 फरवरी, 2016 को कैलिफोर्निया के सांता क्लारा में सैन फ्रांसिस्को खाड़ी क्षेत्र में लेवी स्टेडियम में खेला गया था।</s><2hi>", add_special_tokens=False, return_tensors="pt", padding=True).input_ids 

out = tokenizer("<2hi> सुपर बाउल किस दिन खेला गया? </s>", add_special_tokens=False, return_tensors="pt", padding=True).input_ids 
model_outputs=model(input_ids=inp, decoder_input_ids=out[:,0:-1], labels=out[:,1:])

# For loss
model_outputs.loss ## This is not label smoothed.

# For logits
model_outputs.logits

# For generation. Pardon the messiness. Note the decoder_start_token_id.
model.eval() # Set dropouts to zero

model_output=model.generate(inp, use_cache=True,no_repeat_ngram_size=3,encoder_no_repeat_ngram_size=3, num_beams=4, max_length=20, min_length=1, early_stopping=True, pad_token_id=pad_id, bos_token_id=bos_id, eos_token_id=eos_id, decoder_start_token_id=tokenizer._convert_token_to_id_with_added_voc("<2hi>"))

# Decode to get output strings
decoded_output=tokenizer.decode(model_output[0], skip_special_tokens=True, clean_up_tokenization_spaces=False)
print(decoded_output) # कब खेला जाएगा पहला मैच?

# Disclaimer
Note that if your output language is not Hindi or Marathi, you should convert its script from Devanagari to the desired language using the [Indic NLP Library](https://github.com/AI4Bharat/indic-bart/blob/main/indic_scriptmap.py).

Note:

If you wish to use any language written in a non-Devanagari script, then you should first convert it to Devanagari using the Indic NLP Library. After you get the output, you should convert it back into the original script.

Benchmarks

Scores on the IndicQuestionGeneration test sets are as follows:

Language RougeL
as 20.48
bn 26.63
gu 27.71
hi 35.38
kn 23.56
ml 22.17
mr 23.52
or 25.25
pa 32.10
ta 22.98
te 25.67

Citation

If you use this model, please cite the following paper:

@inproceedings{Kumar2022IndicNLGSM,
  title={IndicNLG Suite: Multilingual Datasets for Diverse NLG Tasks in Indic Languages},
  author={Aman Kumar and Himani Shrotriya and Prachi Sahu and Raj Dabre and Ratish Puduppully and Anoop Kunchukuttan and Amogh Mishra and Mitesh M. Khapra and Pratyush Kumar},
  year={2022},
  url = "https://arxiv.org/abs/2203.05437"
  }

License

The model is available under the MIT License.

Downloads last month
29
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.

Datasets used to train ai4bharat/MultiIndicQuestionGenerationUnified