metadata

license: cc-by-4.0
metrics:
  - bleu4
  - meteor
  - rouge-l
  - bertscore
  - moverscore
language: ko
datasets:
  - lmqg/qag_koquad
pipeline_tag: text2text-generation
tags:
  - questions and answers generation
widget:
  - text: >-
      1990년 영화 《 남부군 》에서 단역으로 영화배우 첫 데뷔에 이어 같은 해 KBS 드라마 《지구인》에서 단역으로 출연하였고 이듬해
      MBC 《여명의 눈동자》를 통해 단역으로 출연하였다.
    example_title: Questions & Answers Generation Example 1
model-index:
  - name: lmqg/mt5-small-koquad-qag
    results:
      - task:
          name: Text2text Generation
          type: text2text-generation
        dataset:
          name: lmqg/qag_koquad
          type: default
          args: default
        metrics:
          - name: BLEU4 (Question & Answer Generation)
            type: bleu4_question_answer_generation
            value: 0.71
          - name: ROUGE-L (Question & Answer Generation)
            type: rouge_l_question_answer_generation
            value: 8.06
          - name: METEOR (Question & Answer Generation)
            type: meteor_question_answer_generation
            value: 12.95
          - name: BERTScore (Question & Answer Generation)
            type: bertscore_question_answer_generation
            value: 58.39
          - name: MoverScore (Question & Answer Generation)
            type: moverscore_question_answer_generation
            value: 58.81
          - name: QAAlignedF1Score-BERTScore (Question & Answer Generation)
            type: qa_aligned_f1_score_bertscore_question_answer_generation
            value: 74.23
          - name: QAAlignedRecall-BERTScore (Question & Answer Generation)
            type: qa_aligned_recall_bertscore_question_answer_generation
            value: 74.2
          - name: QAAlignedPrecision-BERTScore (Question & Answer Generation)
            type: qa_aligned_precision_bertscore_question_answer_generation
            value: 74.29
          - name: QAAlignedF1Score-MoverScore (Question & Answer Generation)
            type: qa_aligned_f1_score_moverscore_question_answer_generation
            value: 75.06
          - name: QAAlignedRecall-MoverScore (Question & Answer Generation)
            type: qa_aligned_recall_moverscore_question_answer_generation
            value: 75.04
          - name: QAAlignedPrecision-MoverScore (Question & Answer Generation)
            type: qa_aligned_precision_moverscore_question_answer_generation
            value: 75.14

Model Card of `lmqg/mt5-small-koquad-qag`

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

Overview

Language model: google/mt5-small
Language: ko
Training data: lmqg/qag_koquad (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="ko", model="lmqg/mt5-small-koquad-qag")

# model prediction
question_answer_pairs = model.generate_qa("1990년 영화 《 남부군 》에서 단역으로 영화배우 첫 데뷔에 이어 같은 해 KBS 드라마 《지구인》에서 단역으로 출연하였고 이듬해 MBC 《여명의 눈동자》를 통해 단역으로 출연하였다.")

With transformers

from transformers import pipeline

pipe = pipeline("text2text-generation", "lmqg/mt5-small-koquad-qag")
output = pipe("1990년 영화 《 남부군 》에서 단역으로 영화배우 첫 데뷔에 이어 같은 해 KBS 드라마 《지구인》에서 단역으로 출연하였고 이듬해 MBC 《여명의 눈동자》를 통해 단역으로 출연하였다.")

Evaluation

Metric (Question & Answer Generation): raw metric file

	Score	Type	Dataset
BERTScore	58.39	default	lmqg/qag_koquad
Bleu_1	4.09	default	lmqg/qag_koquad
Bleu_2	2.09	default	lmqg/qag_koquad
Bleu_3	1.17	default	lmqg/qag_koquad
Bleu_4	0.71	default	lmqg/qag_koquad
METEOR	12.95	default	lmqg/qag_koquad
MoverScore	58.81	default	lmqg/qag_koquad
QAAlignedF1Score (BERTScore)	74.23	default	lmqg/qag_koquad
QAAlignedF1Score (MoverScore)	75.06	default	lmqg/qag_koquad
QAAlignedPrecision (BERTScore)	74.29	default	lmqg/qag_koquad
QAAlignedPrecision (MoverScore)	75.14	default	lmqg/qag_koquad
QAAlignedRecall (BERTScore)	74.2	default	lmqg/qag_koquad
QAAlignedRecall (MoverScore)	75.04	default	lmqg/qag_koquad
ROUGE_L	8.06	default	lmqg/qag_koquad

Training hyperparameters

The following hyperparameters were used during fine-tuning:

dataset_path: lmqg/qag_koquad
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: 13
batch: 8
lr: 0.0005
fp16: False
random_seed: 1
gradient_accumulation_steps: 16
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",
}