Edit model card

Model Card of lmqg/mt5-base-jaquad-qg

This model is fine-tuned version of google/mt5-base for question generation task on the lmqg/qg_jaquad (dataset_name: default) via lmqg.

Overview

Usage

from lmqg import TransformersQG

# initialize model
model = TransformersQG(language="ja", model="lmqg/mt5-base-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/mt5-base-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

Score Type Dataset
BERTScore 81.77 default lmqg/qg_jaquad
Bleu_1 57.89 default lmqg/qg_jaquad
Bleu_2 46.06 default lmqg/qg_jaquad
Bleu_3 38.25 default lmqg/qg_jaquad
Bleu_4 32.54 default lmqg/qg_jaquad
METEOR 30.58 default lmqg/qg_jaquad
MoverScore 59.68 default lmqg/qg_jaquad
ROUGE_L 52.67 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) 86.66 default lmqg/qg_jaquad
QAAlignedF1Score (MoverScore) 62.8 default lmqg/qg_jaquad
QAAlignedPrecision (BERTScore) 86.67 default lmqg/qg_jaquad
QAAlignedPrecision (MoverScore) 62.82 default lmqg/qg_jaquad
QAAlignedRecall (BERTScore) 86.65 default lmqg/qg_jaquad
QAAlignedRecall (MoverScore) 62.78 default lmqg/qg_jaquad
Score Type Dataset
QAAlignedF1Score (BERTScore) 80.31 default lmqg/qg_jaquad
QAAlignedF1Score (MoverScore) 56.36 default lmqg/qg_jaquad
QAAlignedPrecision (BERTScore) 77.14 default lmqg/qg_jaquad
QAAlignedPrecision (MoverScore) 54 default lmqg/qg_jaquad
QAAlignedRecall (BERTScore) 83.89 default lmqg/qg_jaquad
QAAlignedRecall (MoverScore) 59.12 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: google/mt5-base
  • max_length: 512
  • max_length_output: 32
  • epoch: 30
  • batch: 32
  • lr: 0.0001
  • fp16: False
  • random_seed: 1
  • gradient_accumulation_steps: 2
  • label_smoothing: 0.0

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",
}
Downloads last month
26

Dataset used to train lmqg/mt5-base-jaquad-qg

Evaluation results

  • BLEU4 (Question Generation) on lmqg/qg_jaquad
    self-reported
    32.540
  • ROUGE-L (Question Generation) on lmqg/qg_jaquad
    self-reported
    52.670
  • METEOR (Question Generation) on lmqg/qg_jaquad
    self-reported
    30.580
  • BERTScore (Question Generation) on lmqg/qg_jaquad
    self-reported
    81.770
  • MoverScore (Question Generation) on lmqg/qg_jaquad
    self-reported
    59.680
  • QAAlignedF1Score-BERTScore (Question & Answer Generation (with Gold Answer)) [Gold Answer] on lmqg/qg_jaquad
    self-reported
    86.660
  • QAAlignedRecall-BERTScore (Question & Answer Generation (with Gold Answer)) [Gold Answer] on lmqg/qg_jaquad
    self-reported
    86.650
  • QAAlignedPrecision-BERTScore (Question & Answer Generation (with Gold Answer)) [Gold Answer] on lmqg/qg_jaquad
    self-reported
    86.670
  • QAAlignedF1Score-MoverScore (Question & Answer Generation (with Gold Answer)) [Gold Answer] on lmqg/qg_jaquad
    self-reported
    62.800
  • QAAlignedRecall-MoverScore (Question & Answer Generation (with Gold Answer)) [Gold Answer] on lmqg/qg_jaquad
    self-reported
    62.780
  • QAAlignedPrecision-MoverScore (Question & Answer Generation (with Gold Answer)) [Gold Answer] on lmqg/qg_jaquad
    self-reported
    62.820
  • QAAlignedF1Score-BERTScore (Question & Answer Generation) [Gold Answer] on lmqg/qg_jaquad
    self-reported
    80.310
  • QAAlignedRecall-BERTScore (Question & Answer Generation) [Gold Answer] on lmqg/qg_jaquad
    self-reported
    83.890
  • QAAlignedPrecision-BERTScore (Question & Answer Generation) [Gold Answer] on lmqg/qg_jaquad
    self-reported
    77.140
  • QAAlignedF1Score-MoverScore (Question & Answer Generation) [Gold Answer] on lmqg/qg_jaquad
    self-reported
    56.360
  • QAAlignedRecall-MoverScore (Question & Answer Generation) [Gold Answer] on lmqg/qg_jaquad
    self-reported
    59.120
  • QAAlignedPrecision-MoverScore (Question & Answer Generation) [Gold Answer] on lmqg/qg_jaquad
    self-reported
    54.000