Model Card of lmqg/t5-large-tweetqa-qa

This model is fine-tuned version of t5-large for question answering task on the lmqg/qg_tweetqa (dataset_name: default) via lmqg.

Overview

• Language model: t5-large
• Language: en
• Training data: lmqg/qg_tweetqa (default)
• Online Demo: https://autoqg.net/
• Repository: https://github.com/asahi417/lm-question-generation
• Paper: https://arxiv.org/abs/2210.03992

Usage

• With lmqg

from lmqg import TransformersQG

# initialize model
model = TransformersQG(language="en", model="lmqg/t5-large-tweetqa-qa")

# model prediction
answers = model.answer_q(list_question="What is a person called is practicing heresy?", list_context=" Heresy is any provocative belief or theory that is strongly at variance with established beliefs or customs. A heretic is a proponent of such claims or beliefs. Heresy is distinct from both apostasy, which is the explicit renunciation of one's religion, principles or cause, and blasphemy, which is an impious utterance or action concerning God or sacred things.")

• With transformers

from transformers import pipeline

pipe = pipeline("text2text-generation", "lmqg/t5-large-tweetqa-qa")
output = pipe("question: What is a person called is practicing heresy?, context: Heresy is any provocative belief or theory that is strongly at variance with established beliefs or customs. A heretic is a proponent of such claims or beliefs. Heresy is distinct from both apostasy, which is the explicit renunciation of one's religion, principles or cause, and blasphemy, which is an impious utterance or action concerning God or sacred things.")

Evaluation

• Metric (Question Answering): raw metric file

	Score	Type	Dataset
AnswerExactMatch	54.45	default	lmqg/qg_tweetqa
AnswerF1Score	71.1	default	lmqg/qg_tweetqa
BERTScore	94.8	default	lmqg/qg_tweetqa
Bleu_1	58.53	default	lmqg/qg_tweetqa
Bleu_2	49.65	default	lmqg/qg_tweetqa
Bleu_3	41.43	default	lmqg/qg_tweetqa
Bleu_4	35.02	default	lmqg/qg_tweetqa
METEOR	36.5	default	lmqg/qg_tweetqa
MoverScore	80.79	default	lmqg/qg_tweetqa
ROUGE_L	64.13	default	lmqg/qg_tweetqa

Training hyperparameters

The following hyperparameters were used during fine-tuning:

• dataset_path: lmqg/qg_tweetqa
• dataset_name: default
• input_types: ['paragraph_question']
• output_types: ['answer']
• prefix_types: None
• model: t5-large
• max_length: 512
• max_length_output: 32
• epoch: 6
• batch: 16
• lr: 5e-05
• fp16: False
• random_seed: 1
• gradient_accumulation_steps: 4
• label_smoothing: 0.15

The full configuration can be found at fine-tuning config file.

Citation

@inproceedings{ushio-etal-2022-generative,
    title = "{G}enerative {L}anguage {M}odels for {P}aragraph-{L}evel {Q}uestion {G}eneration",
    author = "Ushio, Asahi  and
        Alva-Manchego, Fernando  and
        Camacho-Collados, Jose",
    booktitle = "Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing",
    month = dec,
    year = "2022",
    address = "Abu Dhabi, U.A.E.",
    publisher = "Association for Computational Linguistics",
}