mt5-small-ruquad-qg / README.md
asahi417's picture
model update
663c0cf
|
raw
history blame
10.3 kB
metadata
license: cc-by-4.0
metrics:
  - bleu4
  - meteor
  - rouge-l
  - bertscore
  - moverscore
language: ru
datasets:
  - lmqg/qg_ruquad
pipeline_tag: text2text-generation
tags:
  - question generation
widget:
  - text: >-
      Нелишним будет отметить, что, развивая это направление, Д. И. Менделеев,
      поначалу априорно выдвинув идею о температуре, при которой высота мениска
      будет нулевой, <hl> в мае 1860 года <hl> провёл серию опытов.
    example_title: Question Generation Example 1
  - text: >-
      Однако, франкоязычный <hl> Квебек <hl> практически никогда не включается в
      состав Латинской Америки.
    example_title: Question Generation Example 2
  - text: >-
      Классическим примером международного синдиката XX века была группа
      компаний <hl> Де Бирс <hl> , которая в 1980-е годы контролировала до 90 %
      мировой торговли алмазами.
    example_title: Question Generation Example 3
model-index:
  - name: lmqg/mt5-small-ruquad-qg
    results:
      - task:
          name: Text2text Generation
          type: text2text-generation
        dataset:
          name: lmqg/qg_ruquad
          type: default
          args: default
        metrics:
          - name: BLEU4 (Question Generation)
            type: bleu4_question_generation
            value: 16.31
          - name: ROUGE-L (Question Generation)
            type: rouge_l_question_generation
            value: 31.39
          - name: METEOR (Question Generation)
            type: meteor_question_generation
            value: 26.39
          - name: BERTScore (Question Generation)
            type: bertscore_question_generation
            value: 84.27
          - name: MoverScore (Question Generation)
            type: moverscore_question_generation
            value: 62.49
          - name: BLEU4 (Question & Answer Generation)
            type: bleu4_question_answer_generation
            value: 18.61
          - name: ROUGE-L (Question & Answer Generation)
            type: rouge_l_question_answer_generation
            value: 51.04
          - name: METEOR (Question & Answer Generation)
            type: meteor_question_answer_generation
            value: 43.1
          - name: BERTScore (Question & Answer Generation)
            type: bertscore_question_answer_generation
            value: 90.03
          - name: MoverScore (Question & Answer Generation)
            type: moverscore_question_answer_generation
            value: 67.82
          - name: >-
              QAAlignedF1Score-BERTScore (Question & Answer Generation) [Gold
              Answer]
            type: >-
              qa_aligned_f1_score_bertscore_question_answer_generation_gold_answer
            value: 90.17
          - name: >-
              QAAlignedRecall-BERTScore (Question & Answer Generation) [Gold
              Answer]
            type: qa_aligned_recall_bertscore_question_answer_generation_gold_answer
            value: 90.16
          - name: >-
              QAAlignedPrecision-BERTScore (Question & Answer Generation) [Gold
              Answer]
            type: >-
              qa_aligned_precision_bertscore_question_answer_generation_gold_answer
            value: 90.17
          - name: >-
              QAAlignedF1Score-MoverScore (Question & Answer Generation) [Gold
              Answer]
            type: >-
              qa_aligned_f1_score_moverscore_question_answer_generation_gold_answer
            value: 68.22
          - name: >-
              QAAlignedRecall-MoverScore (Question & Answer Generation) [Gold
              Answer]
            type: >-
              qa_aligned_recall_moverscore_question_answer_generation_gold_answer
            value: 68.21
          - name: >-
              QAAlignedPrecision-MoverScore (Question & Answer Generation) [Gold
              Answer]
            type: >-
              qa_aligned_precision_moverscore_question_answer_generation_gold_answer
            value: 68.23

Model Card of lmqg/mt5-small-ruquad-qg

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

Overview

Usage

from lmqg import TransformersQG

# initialize model
model = TransformersQG(language="ru", model="lmqg/mt5-small-ruquad-qg")

# model prediction
questions = model.generate_q(list_context="Нелишним будет отметить, что, развивая это направление, Д. И. Менделеев, поначалу априорно выдвинув идею о температуре, при которой высота мениска будет нулевой, в мае 1860 года провёл серию опытов.", list_answer="в мае 1860 года")
  • With transformers
from transformers import pipeline

pipe = pipeline("text2text-generation", "lmqg/mt5-small-ruquad-qg")
output = pipe("Нелишним будет отметить, что, развивая это направление, Д. И. Менделеев, поначалу априорно выдвинув идею о температуре, при которой высота мениска будет нулевой, <hl> в мае 1860 года <hl> провёл серию опытов.")

Evaluation

Score Type Dataset
BERTScore 84.27 default lmqg/qg_ruquad
Bleu_1 31.03 default lmqg/qg_ruquad
Bleu_2 24.58 default lmqg/qg_ruquad
Bleu_3 19.92 default lmqg/qg_ruquad
Bleu_4 16.31 default lmqg/qg_ruquad
METEOR 26.39 default lmqg/qg_ruquad
MoverScore 62.49 default lmqg/qg_ruquad
ROUGE_L 31.39 default lmqg/qg_ruquad
  • 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
BERTScore 90.03 default lmqg/qg_ruquad
Bleu_1 45.81 default lmqg/qg_ruquad
Bleu_2 34.13 default lmqg/qg_ruquad
Bleu_3 25.81 default lmqg/qg_ruquad
Bleu_4 18.61 default lmqg/qg_ruquad
METEOR 43.1 default lmqg/qg_ruquad
MoverScore 67.82 default lmqg/qg_ruquad
QAAlignedF1Score (BERTScore) 90.17 default lmqg/qg_ruquad
QAAlignedF1Score (MoverScore) 68.22 default lmqg/qg_ruquad
QAAlignedPrecision (BERTScore) 90.17 default lmqg/qg_ruquad
QAAlignedPrecision (MoverScore) 68.23 default lmqg/qg_ruquad
QAAlignedRecall (BERTScore) 90.16 default lmqg/qg_ruquad
QAAlignedRecall (MoverScore) 68.21 default lmqg/qg_ruquad
ROUGE_L 51.04 default lmqg/qg_ruquad

Training hyperparameters

The following hyperparameters were used during fine-tuning:

  • dataset_path: lmqg/qg_ruquad
  • dataset_name: default
  • input_types: ['paragraph_answer']
  • output_types: ['question']
  • prefix_types: None
  • model: google/mt5-small
  • max_length: 512
  • max_length_output: 32
  • epoch: 5
  • batch: 64
  • lr: 0.001
  • fp16: False
  • random_seed: 1
  • gradient_accumulation_steps: 1
  • 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",
}