Model Card of research-backup/t5-large-tweetqa-qag-np

This model is fine-tuned version of t5-large for question & answer pair generation task on the lmqg/qag_tweetqa (dataset_name: default) via lmqg. This model is fine-tuned without a task prefix.

Overview

• Language model: t5-large
• Language: en
• Training data: lmqg/qag_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="research-backup/t5-large-tweetqa-qag-np")

# model prediction
question_answer_pairs = model.generate_qa("William Turner was an English painter who specialised in watercolour landscapes")

• With transformers

from transformers import pipeline

pipe = pipeline("text2text-generation", "research-backup/t5-large-tweetqa-qag-np")
output = pipe("Beyonce further expanded her acting career, starring as blues singer Etta James in the 2008 musical biopic, Cadillac Records.")

Evaluation

• Metric (Question & Answer Generation): raw metric file

	Score	Type	Dataset
BERTScore	90.95	default	lmqg/qag_tweetqa
Bleu_1	40.9	default	lmqg/qag_tweetqa
Bleu_2	28.27	default	lmqg/qag_tweetqa
Bleu_3	19.84	default	lmqg/qag_tweetqa
Bleu_4	14.14	default	lmqg/qag_tweetqa
METEOR	31.49	default	lmqg/qag_tweetqa
MoverScore	62.62	default	lmqg/qag_tweetqa
QAAlignedF1Score (BERTScore)	92.64	default	lmqg/qag_tweetqa
QAAlignedF1Score (MoverScore)	65.47	default	lmqg/qag_tweetqa
QAAlignedPrecision (BERTScore)	93.03	default	lmqg/qag_tweetqa
QAAlignedPrecision (MoverScore)	66.36	default	lmqg/qag_tweetqa
QAAlignedRecall (BERTScore)	92.27	default	lmqg/qag_tweetqa
QAAlignedRecall (MoverScore)	64.68	default	lmqg/qag_tweetqa
ROUGE_L	37.45	default	lmqg/qag_tweetqa

Training hyperparameters

The following hyperparameters were used during fine-tuning:

• dataset_path: lmqg/qag_tweetqa
• dataset_name: default
• input_types: ['paragraph']
• output_types: ['questions_answers']
• prefix_types: None
• model: t5-large
• max_length: 256
• max_length_output: 128
• epoch: 16
• batch: 16
• lr: 0.0001
• 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",
}