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

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

Overview

• Language model: facebook/mbart-large-cc25
• Language: es
• Training data: lmqg/qg_esquad (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="es", model="lmqg/mbart-large-cc25-esquad-qg")

# model prediction
questions = model.generate_q(list_context="a noviembre , que es también la estación lluviosa.", list_answer="noviembre")

• With transformers

from transformers import pipeline

pipe = pipeline("text2text-generation", "lmqg/mbart-large-cc25-esquad-qg")
output = pipe("del <hl> Ministerio de Desarrollo Urbano <hl> , Gobierno de la India.")

Evaluation

• Metric (Question Generation): raw metric file

	Score	Type	Dataset
BERTScore	83.58	default	lmqg/qg_esquad
Bleu_1	25.02	default	lmqg/qg_esquad
Bleu_2	16.93	default	lmqg/qg_esquad
Bleu_3	12.28	default	lmqg/qg_esquad
Bleu_4	9.18	default	lmqg/qg_esquad
METEOR	22.95	default	lmqg/qg_esquad
MoverScore	58.91	default	lmqg/qg_esquad
ROUGE_L	24.26	default	lmqg/qg_esquad

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

	Score	Type	Dataset
QAAlignedF1Score (BERTScore)	88.99	default	lmqg/qg_esquad
QAAlignedF1Score (MoverScore)	63.47	default	lmqg/qg_esquad
QAAlignedPrecision (BERTScore)	89	default	lmqg/qg_esquad
QAAlignedPrecision (MoverScore)	63.49	default	lmqg/qg_esquad
QAAlignedRecall (BERTScore)	88.97	default	lmqg/qg_esquad
QAAlignedRecall (MoverScore)	63.46	default	lmqg/qg_esquad

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

	Score	Type	Dataset
QAAlignedF1Score (BERTScore)	79.26	default	lmqg/qg_esquad
QAAlignedF1Score (MoverScore)	54.68	default	lmqg/qg_esquad
QAAlignedPrecision (BERTScore)	76.78	default	lmqg/qg_esquad
QAAlignedPrecision (MoverScore)	53.17	default	lmqg/qg_esquad
QAAlignedRecall (BERTScore)	81.98	default	lmqg/qg_esquad
QAAlignedRecall (MoverScore)	56.37	default	lmqg/qg_esquad

Training hyperparameters

The following hyperparameters were used during fine-tuning:

• dataset_path: lmqg/qg_esquad
• 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: 4
• 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",
}