Edit model card

t5-large fine-tuned to RACE for Generating Question+Answer

  • Input: context (e.g. news article)
  • Output: question <sep> answer

This model generates abstractive answers following the RACE dataset. If you would like to have extractive questions/answers, you can use our model trained on SQuAD: https://huggingface.co/potsawee/t5-large-generation-squad-QuestionAnswer.

Model Details

t5-large model is fine-tuned to the RACE dataset where the input is the context/passage and the output is the question followed by the answer. This is the first component in the question generation pipeline (i.e. g1) in our MQAG paper, or please refer to the GitHub repo of this project: https://github.com/potsawee/mqag0.

How to Use the Model

Use the code below to get started with the model. You can also set do_sample=True in generate() to obtain different question-answer pairs.

>>> from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

>>> tokenizer = AutoTokenizer.from_pretrained("potsawee/t5-large-generation-race-QuestionAnswer")
>>> model = AutoModelForSeq2SeqLM.from_pretrained("potsawee/t5-large-generation-race-QuestionAnswer")

>>> context = r"""
... World number one Novak Djokovic says he is hoping for a "positive decision" to allow him 
... to play at Indian Wells and the Miami Open next month. The United States has extended 
... its requirement for international visitors to be vaccinated against Covid-19. Proof of vaccination 
... will be required to enter the country until at least 10 April, but the Serbian has previously 
... said he is unvaccinated. The 35-year-old has applied for special permission to enter the country. 
... Indian Wells and the Miami Open - two of the most prestigious tournaments on the tennis calendar 
... outside the Grand Slams - start on 6 and 20 March respectively. Djokovic says he will return to 
... the ATP tour in Dubai next week after claiming a record-extending 10th Australian Open title 
... and a record-equalling 22nd Grand Slam men's title last month.""".replace("\n", "")

>>> inputs = tokenizer(context, return_tensors="pt")
>>> outputs = model.generate(**inputs, max_length=100)
>>> question_answer = tokenizer.decode(outputs[0], skip_special_tokens=False)
>>> question_answer = question_answer.replace(tokenizer.pad_token, "").replace(tokenizer.eos_token, "")
>>> question, answer = question_answer.split(tokenizer.sep_token)

>>> print("question:", question)
question:  What is the best title for the passage?
>>> print("answer:", answer)
answer:  Djokovic's application for special permission to enter the United States

Generating Distractors (other options in a multiple-choice setup)

Context ---> Question + (A) Answer (B) Distractor1 (C) Distractor2 (D) Distractor3

Please refer to our distractor generation model: https://huggingface.co/potsawee/t5-large-generation-race-Distractor

Citation

@article{manakul2023mqag,
  title={MQAG: Multiple-choice Question Answering and Generation for Assessing Information Consistency in Summarization},
  author={Manakul, Potsawee and Liusie, Adian and Gales, Mark JF},
  journal={arXiv preprint arXiv:2301.12307},
  year={2023}
}
Downloads last month
276

Dataset used to train potsawee/t5-large-generation-race-QuestionAnswer