metadata

license: cc-by-4.0
metrics:
  - bleu4
  - meteor
  - rouge-l
  - bertscore
  - moverscore
language: de
datasets:
  - lmqg/qg_dequad
pipeline_tag: text2text-generation
tags:
  - question generation
widget:
  - text: >-
      Empfangs- und Sendeantenne sollen in ihrer Polarisation übereinstimmen,
      andernfalls <hl> wird die Signalübertragung stark gedämpft. <hl>
    example_title: Question Generation Example 1
  - text: >-
      das erste weltweit errichtete Hermann Brehmer <hl> 1855 <hl> im
      niederschlesischen ''Görbersdorf'' (heute Sokołowsko, Polen).
    example_title: Question Generation Example 2
  - text: >-
      Er muss Zyperngrieche sein und wird direkt für <hl> fünf Jahre <hl>
      gewählt (Art. 43 Abs. 1 der Verfassung) und verfügt über weitreichende
      Exekutivkompetenzen.
    example_title: Question Generation Example 3
model-index:
  - name: lmqg/mbart-large-cc25-dequad-qg
    results:
      - task:
          name: Text2text Generation
          type: text2text-generation
        dataset:
          name: lmqg/qg_dequad
          type: default
          args: default
        metrics:
          - name: BLEU4 (Question Generation)
            type: bleu4_question_generation
            value: 0.75
          - name: ROUGE-L (Question Generation)
            type: rouge_l_question_generation
            value: 11.19
          - name: METEOR (Question Generation)
            type: meteor_question_generation
            value: 13.71
          - name: BERTScore (Question Generation)
            type: bertscore_question_generation
            value: 80.77
          - name: MoverScore (Question Generation)
            type: moverscore_question_generation
            value: 55.88
          - name: QAAlignedF1Score-BERTScore (Gold Answer)
            type: qa_aligned_f1_score_bertscore_gold_answer
            value: 90.66
          - name: QAAlignedRecall-BERTScore (Gold Answer)
            type: qa_aligned_recall_bertscore_gold_answer
            value: 90.69
          - name: QAAlignedPrecision-BERTScore (Gold Answer)
            type: qa_aligned_precision_bertscore_gold_answer
            value: 90.64
          - name: QAAlignedF1Score-MoverScore (Gold Answer)
            type: qa_aligned_f1_score_moverscore_gold_answer
            value: 65.36
          - name: QAAlignedRecall-MoverScore (Gold Answer)
            type: qa_aligned_recall_moverscore_gold_answer
            value: 65.36
          - name: QAAlignedPrecision-MoverScore (Gold Answer)
            type: qa_aligned_precision_moverscore_gold_answer
            value: 65.37

Model Card of `lmqg/mbart-large-cc25-dequad-qg`

This model is fine-tuned version of facebook/mbart-large-cc25 for question generation task on the lmqg/qg_dequad (dataset_name: default) via lmqg.

Overview

Language model: facebook/mbart-large-cc25
Language: de
Training data: lmqg/qg_dequad (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="de", model="lmqg/mbart-large-cc25-dequad-qg")

# model prediction
questions = model.generate_q(list_context="das erste weltweit errichtete Hermann Brehmer 1855 im niederschlesischen ''Görbersdorf'' (heute Sokołowsko, Polen).", list_answer="1855")

With transformers

from transformers import pipeline

pipe = pipeline("text2text-generation", "lmqg/mbart-large-cc25-dequad-qg")
output = pipe("Empfangs- und Sendeantenne sollen in ihrer Polarisation übereinstimmen, andernfalls <hl> wird die Signalübertragung stark gedämpft. <hl>")

Evaluation

Metric (Question Generation): raw metric file

	Score	Type	Dataset
BERTScore	80.77	default	lmqg/qg_dequad
Bleu_1	10.96	default	lmqg/qg_dequad
Bleu_2	4.48	default	lmqg/qg_dequad
Bleu_3	1.91	default	lmqg/qg_dequad
Bleu_4	0.75	default	lmqg/qg_dequad
METEOR	13.71	default	lmqg/qg_dequad
MoverScore	55.88	default	lmqg/qg_dequad
ROUGE_L	11.19	default	lmqg/qg_dequad

Metric (Question & Answer Generation): QAG metrics are computed with the gold answer and generated question on it for this model, as the model cannot provide an answer. raw metric file

	Score	Type	Dataset
QAAlignedF1Score (BERTScore)	90.66	default	lmqg/qg_dequad
QAAlignedF1Score (MoverScore)	65.36	default	lmqg/qg_dequad
QAAlignedPrecision (BERTScore)	90.64	default	lmqg/qg_dequad
QAAlignedPrecision (MoverScore)	65.37	default	lmqg/qg_dequad
QAAlignedRecall (BERTScore)	90.69	default	lmqg/qg_dequad
QAAlignedRecall (MoverScore)	65.36	default	lmqg/qg_dequad

Training hyperparameters

The following hyperparameters were used during fine-tuning:

dataset_path: lmqg/qg_dequad
dataset_name: default
input_types: ['paragraph_answer']
output_types: ['question']
prefix_types: None
model: facebook/mbart-large-cc25
max_length: 512
max_length_output: 32
epoch: 11
batch: 4
lr: 0.0001
fp16: False
random_seed: 1
gradient_accumulation_steps: 16
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",
}