Model Card of lmqg/mbart-large-cc25-jaquad-qg

This model is fine-tuned version of facebook/mbart-large-cc25 for question generation task on the lmqg/qg_jaquad (dataset_name: default) via lmqg.

Overview

• Language model: facebook/mbart-large-cc25
• Language: ja
• Training data: lmqg/qg_jaquad (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="ja", model="lmqg/mbart-large-cc25-jaquad-qg")

# model prediction
questions = model.generate_q(list_context="フェルメールの作品では、17世紀のオランダの画家、ヨハネス・フェルメールの作品について記述する。フェルメールの作品は、疑問作も含め30数点しか現存しない。現存作品はすべて油彩画で、版画、下絵、素描などは残っていない。", list_answer="30数点")

• With transformers

from transformers import pipeline

pipe = pipeline("text2text-generation", "lmqg/mbart-large-cc25-jaquad-qg")
output = pipe("ゾフィーは貴族出身ではあったが王族出身ではなく、ハプスブルク家の皇位継承者であるフランツ・フェルディナントとの結婚は貴賤結婚となった。皇帝フランツ・ヨーゼフは、2人の間に生まれた子孫が皇位を継がないことを条件として結婚を承認していた。視察が予定されている<hl>6月28日<hl>は2人の14回目の結婚記念日であった。")

Evaluation

• Metric (Question Generation): raw metric file

	Score	Type	Dataset
BERTScore	82.26	default	lmqg/qg_jaquad
Bleu_1	57.05	default	lmqg/qg_jaquad
Bleu_2	45.45	default	lmqg/qg_jaquad
Bleu_3	37.81	default	lmqg/qg_jaquad
Bleu_4	32.16	default	lmqg/qg_jaquad
METEOR	29.97	default	lmqg/qg_jaquad
MoverScore	59.88	default	lmqg/qg_jaquad
ROUGE_L	52.95	default	lmqg/qg_jaquad

• Metric (Question & Answer Generation, Reference Answer): Each question is generated from the gold answer. raw metric file

	Score	Type	Dataset
QAAlignedF1Score (BERTScore)	87.16	default	lmqg/qg_jaquad
QAAlignedF1Score (MoverScore)	63.08	default	lmqg/qg_jaquad
QAAlignedPrecision (BERTScore)	87.17	default	lmqg/qg_jaquad
QAAlignedPrecision (MoverScore)	63.1	default	lmqg/qg_jaquad
QAAlignedRecall (BERTScore)	87.16	default	lmqg/qg_jaquad
QAAlignedRecall (MoverScore)	63.06	default	lmqg/qg_jaquad

• Metric (Question & Answer Generation, Pipeline Approach): Each question is generated on the answer generated by lmqg/mbart-large-cc25-jaquad-ae. raw metric file

	Score	Type	Dataset
QAAlignedF1Score (BERTScore)	76.7	default	lmqg/qg_jaquad
QAAlignedF1Score (MoverScore)	53.64	default	lmqg/qg_jaquad
QAAlignedPrecision (BERTScore)	74.75	default	lmqg/qg_jaquad
QAAlignedPrecision (MoverScore)	52.32	default	lmqg/qg_jaquad
QAAlignedRecall (BERTScore)	78.87	default	lmqg/qg_jaquad
QAAlignedRecall (MoverScore)	55.14	default	lmqg/qg_jaquad

Training hyperparameters

The following hyperparameters were used during fine-tuning:

• dataset_path: lmqg/qg_jaquad
• dataset_name: default
• input_types: ['paragraph_answer']
• output_types: ['question']
• prefix_types: None
• model: facebook/mbart-large-cc25
• max_length: 512
• max_length_output: 32
• epoch: 12
• batch: 64
• lr: 0.0001
• fp16: False
• random_seed: 1
• gradient_accumulation_steps: 1
• 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",
}