mt5-base-koquad-qg / README.md
asahi417's picture
model update
3566fbf
|
raw
history blame
10.2 kB
---
license: cc-by-4.0
metrics:
- bleu4
- meteor
- rouge-l
- bertscore
- moverscore
language: ko
datasets:
- lmqg/qg_koquad
pipeline_tag: text2text-generation
tags:
- question generation
widget:
- text: "1990년 영화 《 <hl> 남부군 <hl> 》에서 단역으로 영화배우 첫 데뷔에 이어 같은 해 KBS 드라마 《지구인》에서 단역으로 출연하였고 이듬해 MBC 《여명의 눈동자》를 통해 단역으로 출연하였다."
example_title: "Question Generation Example 1"
- text: "백신이 없기때문에 예방책은 <hl> 살충제 <hl> 를 사용하면서 서식 장소(찻찬 받침, 배수로, 고인 물의 열린 저장소, 버려진 타이어 등)의 수를 줄임으로써 매개체를 통제할 수 있다."
example_title: "Question Generation Example 2"
- text: "<hl> 원테이크 촬영 <hl> 이기 때문에 한 사람이 실수를 하면 처음부터 다시 찍어야 하는 상황이 발생한다."
example_title: "Question Generation Example 3"
model-index:
- name: lmqg/mt5-base-koquad-qg
results:
- task:
name: Text2text Generation
type: text2text-generation
dataset:
name: lmqg/qg_koquad
type: default
args: default
metrics:
- name: BLEU4 (Question Generation)
type: bleu4_question_generation
value: 12.18
- name: ROUGE-L (Question Generation)
type: rouge_l_question_generation
value: 28.57
- name: METEOR (Question Generation)
type: meteor_question_generation
value: 29.62
- name: BERTScore (Question Generation)
type: bertscore_question_generation
value: 84.52
- name: MoverScore (Question Generation)
type: moverscore_question_generation
value: 83.36
- name: BLEU4 (Question & Answer Generation (with Gold Answer))
type: bleu4_question_answer_generation_with_gold_answer
value: 12.08
- name: ROUGE-L (Question & Answer Generation (with Gold Answer))
type: rouge_l_question_answer_generation_with_gold_answer
value: 48.7
- name: METEOR (Question & Answer Generation (with Gold Answer))
type: meteor_question_answer_generation_with_gold_answer
value: 38.75
- name: BERTScore (Question & Answer Generation (with Gold Answer))
type: bertscore_question_answer_generation_with_gold_answer
value: 88.5
- name: MoverScore (Question & Answer Generation (with Gold Answer))
type: moverscore_question_answer_generation_with_gold_answer
value: 85.18
- name: QAAlignedF1Score-BERTScore (Question & Answer Generation (with Gold Answer)) [Gold Answer]
type: qa_aligned_f1_score_bertscore_question_answer_generation_with_gold_answer_gold_answer
value: 88.8
- name: QAAlignedRecall-BERTScore (Question & Answer Generation (with Gold Answer)) [Gold Answer]
type: qa_aligned_recall_bertscore_question_answer_generation_with_gold_answer_gold_answer
value: 88.76
- name: QAAlignedPrecision-BERTScore (Question & Answer Generation (with Gold Answer)) [Gold Answer]
type: qa_aligned_precision_bertscore_question_answer_generation_with_gold_answer_gold_answer
value: 88.84
- name: QAAlignedF1Score-MoverScore (Question & Answer Generation (with Gold Answer)) [Gold Answer]
type: qa_aligned_f1_score_moverscore_question_answer_generation_with_gold_answer_gold_answer
value: 85.93
- name: QAAlignedRecall-MoverScore (Question & Answer Generation (with Gold Answer)) [Gold Answer]
type: qa_aligned_recall_moverscore_question_answer_generation_with_gold_answer_gold_answer
value: 85.87
- name: QAAlignedPrecision-MoverScore (Question & Answer Generation (with Gold Answer)) [Gold Answer]
type: qa_aligned_precision_moverscore_question_answer_generation_with_gold_answer_gold_answer
value: 86.01
---
# Model Card of `lmqg/mt5-base-koquad-qg`
This model is fine-tuned version of [google/mt5-base](https://huggingface.co./google/mt5-base) for question generation task on the [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) (dataset_name: default) via [`lmqg`](https://github.com/asahi417/lm-question-generation).
### Overview
- **Language model:** [google/mt5-base](https://huggingface.co./google/mt5-base)
- **Language:** ko
- **Training data:** [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) (default)
- **Online Demo:** [https://autoqg.net/](https://autoqg.net/)
- **Repository:** [https://github.com/asahi417/lm-question-generation](https://github.com/asahi417/lm-question-generation)
- **Paper:** [https://arxiv.org/abs/2210.03992](https://arxiv.org/abs/2210.03992)
### Usage
- With [`lmqg`](https://github.com/asahi417/lm-question-generation#lmqg-language-model-for-question-generation-)
```python
from lmqg import TransformersQG
# initialize model
model = TransformersQG(language="ko", model="lmqg/mt5-base-koquad-qg")
# model prediction
questions = model.generate_q(list_context="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.", list_answer="남부군")
```
- With `transformers`
```python
from transformers import pipeline
pipe = pipeline("text2text-generation", "lmqg/mt5-base-koquad-qg")
output = pipe("1990년 영화 《 <hl> 남부군 <hl> 》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 Generation)***: [raw metric file](https://huggingface.co./lmqg/mt5-base-koquad-qg/raw/main/eval/metric.first.sentence.paragraph_answer.question.lmqg_qg_koquad.default.json)
| | Score | Type | Dataset |
|:-----------|--------:|:--------|:-----------------------------------------------------------------|
| BERTScore | 84.52 | default | [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) |
| Bleu_1 | 28.54 | default | [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) |
| Bleu_2 | 21.05 | default | [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) |
| Bleu_3 | 15.92 | default | [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) |
| Bleu_4 | 12.18 | default | [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) |
| METEOR | 29.62 | default | [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) |
| MoverScore | 83.36 | default | [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) |
| ROUGE_L | 28.57 | default | [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) |
- ***Metric (Question & Answer Generation, Reference Answer)***: Each question is generated from *the gold answer*. [raw metric file](https://huggingface.co./lmqg/mt5-base-koquad-qg/raw/main/eval/metric.first.answer.paragraph.questions_answers.lmqg_qg_koquad.default.json)
| | Score | Type | Dataset |
|:--------------------------------|--------:|:--------|:-----------------------------------------------------------------|
| BERTScore | 88.5 | default | [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) |
| Bleu_1 | 43.95 | default | [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) |
| Bleu_2 | 32.47 | default | [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) |
| Bleu_3 | 20.34 | default | [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) |
| Bleu_4 | 12.08 | default | [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) |
| METEOR | 38.75 | default | [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) |
| MoverScore | 85.18 | default | [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) |
| QAAlignedF1Score (BERTScore) | 88.8 | default | [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) |
| QAAlignedF1Score (MoverScore) | 85.93 | default | [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) |
| QAAlignedPrecision (BERTScore) | 88.84 | default | [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) |
| QAAlignedPrecision (MoverScore) | 86.01 | default | [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) |
| QAAlignedRecall (BERTScore) | 88.76 | default | [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) |
| QAAlignedRecall (MoverScore) | 85.87 | default | [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) |
| ROUGE_L | 48.7 | default | [lmqg/qg_koquad](https://huggingface.co./datasets/lmqg/qg_koquad) |
## Training hyperparameters
The following hyperparameters were used during fine-tuning:
- dataset_path: lmqg/qg_koquad
- 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: 11
- batch: 4
- lr: 0.0005
- fp16: False
- random_seed: 1
- gradient_accumulation_steps: 16
- label_smoothing: 0.15
The full configuration can be found at [fine-tuning config file](https://huggingface.co./lmqg/mt5-base-koquad-qg/raw/main/trainer_config.json).
## 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",
}
```