Edit model card

Model Card of lmqg/mt5-small-jaquad-qag

This model is fine-tuned version of google/mt5-small for question & answer pair generation task on the lmqg/qag_jaquad (dataset_name: default) via lmqg.

Overview

Usage

from lmqg import TransformersQG

# initialize model
model = TransformersQG(language="ja", model="lmqg/mt5-small-jaquad-qag")

# model prediction
question_answer_pairs = model.generate_qa("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。")
  • With transformers
from transformers import pipeline

pipe = pipeline("text2text-generation", "lmqg/mt5-small-jaquad-qag")
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6月28日は2人の14回目の結婚記念日であった。")

Evaluation

Score Type Dataset
QAAlignedF1Score (BERTScore) 58.35 default lmqg/qag_jaquad
QAAlignedF1Score (MoverScore) 39.19 default lmqg/qag_jaquad
QAAlignedPrecision (BERTScore) 58.34 default lmqg/qag_jaquad
QAAlignedPrecision (MoverScore) 39.21 default lmqg/qag_jaquad
QAAlignedRecall (BERTScore) 58.38 default lmqg/qag_jaquad
QAAlignedRecall (MoverScore) 39.17 default lmqg/qag_jaquad

Training hyperparameters

The following hyperparameters were used during fine-tuning:

  • dataset_path: lmqg/qag_jaquad
  • dataset_name: default
  • input_types: ['paragraph']
  • output_types: ['questions_answers']
  • prefix_types: None
  • model: google/mt5-small
  • max_length: 512
  • max_length_output: 256
  • epoch: 18
  • batch: 8
  • lr: 0.001
  • fp16: False
  • random_seed: 1
  • gradient_accumulation_steps: 8
  • 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
1

Dataset used to train lmqg/mt5-small-jaquad-qag

Evaluation results

  • QAAlignedF1Score-BERTScore (Question & Answer Generation) on lmqg/qag_jaquad
    self-reported
    58.350
  • QAAlignedRecall-BERTScore (Question & Answer Generation) on lmqg/qag_jaquad
    self-reported
    58.380
  • QAAlignedPrecision-BERTScore (Question & Answer Generation) on lmqg/qag_jaquad
    self-reported
    58.340
  • QAAlignedF1Score-MoverScore (Question & Answer Generation) on lmqg/qag_jaquad
    self-reported
    39.190
  • QAAlignedRecall-MoverScore (Question & Answer Generation) on lmqg/qag_jaquad
    self-reported
    39.170
  • QAAlignedPrecision-MoverScore (Question & Answer Generation) on lmqg/qag_jaquad
    self-reported
    39.210