Model Card of lmqg/mt5-small-zhquad-qg-ae
This model is fine-tuned version of google/mt5-small for question generation and answer extraction jointly on the lmqg/qg_zhquad (dataset_name: default) via lmqg
.
Overview
- Language model: google/mt5-small
- Language: zh
- Training data: lmqg/qg_zhquad (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="zh", model="lmqg/mt5-small-zhquad-qg-ae")
# model prediction
question_answer_pairs = model.generate_qa("南安普敦的警察服务由汉普郡警察提供。南安普敦行动的主要基地是一座新的八层专用建筑,造价3000万英镑。该建筑位于南路,2011年启用,靠近南安普敦中央火车站。此前,南安普顿市中心的行动位于市民中心西翼,但由于设施老化,加上计划在旧警察局和地方法院建造一座新博物馆,因此必须搬迁。在Portswood、Banister Park、Hille和Shirley还有其他警察局,在南安普顿中央火车站还有一个英国交通警察局。")
- With
transformers
from transformers import pipeline
pipe = pipeline("text2text-generation", "lmqg/mt5-small-zhquad-qg-ae")
# answer extraction
answer = pipe("generate question: 南安普敦的警察服务由汉普郡警察提供。南安普敦行动的主要基地是一座新的八层专用建筑,造价3000万英镑。该建筑位于南路,2011年启用,靠近<hl> 南安普敦中央 <hl>火车站。此前,南安普顿市中心的行动位于市民中心西翼,但由于设施老化,加上计划在旧警察局和地方法院建造一座新博物馆,因此必须搬迁。在Portswood、Banister Park、Hille和Shirley还有其他警察局,在南安普顿中央火车站还有一个英国交通警察局。")
# question generation
question = pipe("extract answers: 南安普敦的警察服务由汉普郡警察提供。 南安普敦行动的主要基地是一座新的八层专用建筑,造价3000万英镑。 <hl> 该建筑位于南路,2011年启用,靠近 南安普敦中央 火车站。 <hl> 此前,南安普顿市中心的行动位于市民中心西翼,但由于设施老化,加上计划在旧警察局和地方法院建造一座新博物馆,因此必须搬迁。 在Portswood、Banister Park、Hille和Shirley还有其他警察局,在南安普顿中央火车站还有一个英国交通警察局。")
Evaluation
- Metric (Question Generation): raw metric file
Score | Type | Dataset | |
---|---|---|---|
BERTScore | 76.64 | default | lmqg/qg_zhquad |
Bleu_1 | 35.24 | default | lmqg/qg_zhquad |
Bleu_2 | 24.56 | default | lmqg/qg_zhquad |
Bleu_3 | 18.21 | default | lmqg/qg_zhquad |
Bleu_4 | 13.98 | default | lmqg/qg_zhquad |
METEOR | 22.88 | default | lmqg/qg_zhquad |
MoverScore | 57.03 | default | lmqg/qg_zhquad |
ROUGE_L | 33.17 | default | lmqg/qg_zhquad |
- Metric (Question & Answer Generation): raw metric file
Score | Type | Dataset | |
---|---|---|---|
QAAlignedF1Score (BERTScore) | 78.55 | default | lmqg/qg_zhquad |
QAAlignedF1Score (MoverScore) | 53.47 | default | lmqg/qg_zhquad |
QAAlignedPrecision (BERTScore) | 75.41 | default | lmqg/qg_zhquad |
QAAlignedPrecision (MoverScore) | 51.5 | default | lmqg/qg_zhquad |
QAAlignedRecall (BERTScore) | 82.09 | default | lmqg/qg_zhquad |
QAAlignedRecall (MoverScore) | 55.73 | default | lmqg/qg_zhquad |
- Metric (Answer Extraction): raw metric file
Score | Type | Dataset | |
---|---|---|---|
AnswerExactMatch | 93.5 | default | lmqg/qg_zhquad |
AnswerF1Score | 93.58 | default | lmqg/qg_zhquad |
BERTScore | 99.69 | default | lmqg/qg_zhquad |
Bleu_1 | 92 | default | lmqg/qg_zhquad |
Bleu_2 | 88.87 | default | lmqg/qg_zhquad |
Bleu_3 | 85.52 | default | lmqg/qg_zhquad |
Bleu_4 | 81.9 | default | lmqg/qg_zhquad |
METEOR | 69.99 | default | lmqg/qg_zhquad |
MoverScore | 98.34 | default | lmqg/qg_zhquad |
ROUGE_L | 95.05 | default | lmqg/qg_zhquad |
Training hyperparameters
The following hyperparameters were used during fine-tuning:
- dataset_path: lmqg/qg_zhquad
- dataset_name: default
- input_types: ['paragraph_answer', 'paragraph_sentence']
- output_types: ['question', 'answer']
- prefix_types: ['qg', 'ae']
- model: google/mt5-small
- max_length: 512
- max_length_output: 32
- epoch: 13
- batch: 16
- lr: 0.0005
- fp16: False
- random_seed: 1
- gradient_accumulation_steps: 4
- 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",
}
- Downloads last month
- 8
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social
visibility and check back later, or deploy to Inference Endpoints (dedicated)
instead.
Dataset used to train lmqg/mt5-small-zhquad-qg-ae
Evaluation results
- BLEU4 (Question Generation) on lmqg/qg_zhquadself-reported13.980
- ROUGE-L (Question Generation) on lmqg/qg_zhquadself-reported33.170
- METEOR (Question Generation) on lmqg/qg_zhquadself-reported22.880
- BERTScore (Question Generation) on lmqg/qg_zhquadself-reported76.640
- MoverScore (Question Generation) on lmqg/qg_zhquadself-reported57.030
- QAAlignedF1Score-BERTScore (Question & Answer Generation (with Gold Answer)) on lmqg/qg_zhquadself-reported78.550
- QAAlignedRecall-BERTScore (Question & Answer Generation (with Gold Answer)) on lmqg/qg_zhquadself-reported82.090
- QAAlignedPrecision-BERTScore (Question & Answer Generation (with Gold Answer)) on lmqg/qg_zhquadself-reported75.410
- QAAlignedF1Score-MoverScore (Question & Answer Generation (with Gold Answer)) on lmqg/qg_zhquadself-reported53.470
- QAAlignedRecall-MoverScore (Question & Answer Generation (with Gold Answer)) on lmqg/qg_zhquadself-reported55.730