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

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

Overview

• Language model: facebook/mbart-large-cc25
• Language: it
• Training data: lmqg/qg_itquad (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="it", model="lmqg/mbart-large-cc25-itquad-qg")

# model prediction
questions = model.generate_q(list_context="Dopo il 1971 , l' OPEC ha tardato ad adeguare i prezzi per riflettere tale deprezzamento.", list_answer="Dopo il 1971")

• With transformers

from transformers import pipeline

pipe = pipeline("text2text-generation", "lmqg/mbart-large-cc25-itquad-qg")
output = pipe("<hl> Dopo il 1971 <hl> , l' OPEC ha tardato ad adeguare i prezzi per riflettere tale deprezzamento.")

Evaluation

• Metric (Question Generation): raw metric file

	Score	Type	Dataset
BERTScore	80.63	default	lmqg/qg_itquad
Bleu_1	22.51	default	lmqg/qg_itquad
Bleu_2	14.62	default	lmqg/qg_itquad
Bleu_3	10.06	default	lmqg/qg_itquad
Bleu_4	7.13	default	lmqg/qg_itquad
METEOR	17.97	default	lmqg/qg_itquad
MoverScore	56.84	default	lmqg/qg_itquad
ROUGE_L	21.69	default	lmqg/qg_itquad

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

	Score	Type	Dataset
QAAlignedF1Score (BERTScore)	87.56	default	lmqg/qg_itquad
QAAlignedF1Score (MoverScore)	61.71	default	lmqg/qg_itquad
QAAlignedPrecision (BERTScore)	87.62	default	lmqg/qg_itquad
QAAlignedPrecision (MoverScore)	61.83	default	lmqg/qg_itquad
QAAlignedRecall (BERTScore)	87.5	default	lmqg/qg_itquad
QAAlignedRecall (MoverScore)	61.59	default	lmqg/qg_itquad

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

	Score	Type	Dataset
QAAlignedF1Score (BERTScore)	40.13	default	lmqg/qg_itquad
QAAlignedF1Score (MoverScore)	27.8	default	lmqg/qg_itquad
QAAlignedPrecision (BERTScore)	40.43	default	lmqg/qg_itquad
QAAlignedPrecision (MoverScore)	28.09	default	lmqg/qg_itquad
QAAlignedRecall (BERTScore)	39.88	default	lmqg/qg_itquad
QAAlignedRecall (MoverScore)	27.54	default	lmqg/qg_itquad

Training hyperparameters

The following hyperparameters were used during fine-tuning:

• dataset_path: lmqg/qg_itquad
• 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: 8
• batch: 4
• lr: 0.0001
• fp16: False
• random_seed: 1
• gradient_accumulation_steps: 16
• 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",
}