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

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

Overview

• Language model: facebook/mbart-large-cc25
• Language: ru
• Training data: lmqg/qg_ruquad (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="ru", model="lmqg/mbart-large-cc25-ruquad-qg")

# model prediction
questions = model.generate_q(list_context="Нелишним будет отметить, что, развивая это направление, Д. И. Менделеев, поначалу априорно выдвинув идею о температуре, при которой высота мениска будет нулевой, в мае 1860 года провёл серию опытов.", list_answer="в мае 1860 года")

• With transformers

from transformers import pipeline

pipe = pipeline("text2text-generation", "lmqg/mbart-large-cc25-ruquad-qg")
output = pipe("Нелишним будет отметить, что, развивая это направление, Д. И. Менделеев, поначалу априорно выдвинув идею о температуре, при которой высота мениска будет нулевой, <hl> в мае 1860 года <hl> провёл серию опытов.")

Evaluation

• Metric (Question Generation): raw metric file

	Score	Type	Dataset
BERTScore	87.18	default	lmqg/qg_ruquad
Bleu_1	35.25	default	lmqg/qg_ruquad
Bleu_2	28.1	default	lmqg/qg_ruquad
Bleu_3	22.87	default	lmqg/qg_ruquad
Bleu_4	18.8	default	lmqg/qg_ruquad
METEOR	29.3	default	lmqg/qg_ruquad
MoverScore	65.88	default	lmqg/qg_ruquad
ROUGE_L	34.18	default	lmqg/qg_ruquad

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

	Score	Type	Dataset
QAAlignedF1Score (BERTScore)	92.08	default	lmqg/qg_ruquad
QAAlignedF1Score (MoverScore)	71.45	default	lmqg/qg_ruquad
QAAlignedPrecision (BERTScore)	92.09	default	lmqg/qg_ruquad
QAAlignedPrecision (MoverScore)	71.46	default	lmqg/qg_ruquad
QAAlignedRecall (BERTScore)	92.08	default	lmqg/qg_ruquad
QAAlignedRecall (MoverScore)	71.45	default	lmqg/qg_ruquad

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

	Score	Type	Dataset
QAAlignedF1Score (BERTScore)	79.14	default	lmqg/qg_ruquad
QAAlignedF1Score (MoverScore)	56.25	default	lmqg/qg_ruquad
QAAlignedPrecision (BERTScore)	75.88	default	lmqg/qg_ruquad
QAAlignedPrecision (MoverScore)	54.01	default	lmqg/qg_ruquad
QAAlignedRecall (BERTScore)	82.85	default	lmqg/qg_ruquad
QAAlignedRecall (MoverScore)	58.93	default	lmqg/qg_ruquad

Training hyperparameters

The following hyperparameters were used during fine-tuning:

• dataset_path: lmqg/qg_ruquad
• 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: 17
• 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",
}