ArgMining2022-ImageArg

Repo for our paper: ImageArg: A Multi-modal Tweet Dataset for Image Persuasiveness Mining

Annotated Data

data/gun_control.json

Run experiments

Please download tweet content using TwitterAPI into the root directory data before run the experiments. The following parameters need to config to run specific experiments. Please check the code for details.

--data-dir', default='./data', help='path to data'
--exp-dir', default='./experiments/debug', help='path save experimental results'
--exp-mode', default=5, choices=[0,1,2,3,4,5,6], type=int, help='0:persuasive; 1:image content; 2: stance; 3: perusasion_mode; 4:persuasive mode logos; 5:persuasive mode panthos; 6:persuasive mode ethos'
--num-epochs', default=10, type=int, help='number of running epochs'
--data-mode', default=2, choices=[0,1,2], type=int, help='0:text; 1:image; 2:image+text'
--gpus', default='0', type=str, help='specified gpus'
--seed', default=22, type=int, help='random seed number'
--batch-size', default=16, type=int, help='number of samples per batch'
--lr', default=0.001, type=float, help='learning rate'
--persuasive-label-threshold', default=0.6, type=float, help='threshold to categorize persuasive labels'
--kfold', default=5, help='number of fold validation'
--img-model', default=0, choices=[0,1,2], type=int, help='0:Resnet50; 1:Resnet101; 2:VGG16'
--save-checkpoint', default=0, choices=[0,1], type=int, help='0:do not save checkpoints; 1:save checkpoints'
--skip-non-persuasion-mode', default=0, choices=[0,1], type=int, help='0:do not skip; 1:skip'

• Run image modality experiments

python main_image.py \
  --num-epochs=3  \
  --batch-size=16  \
  --exp-mode=0 \
  --data-mode=1 \
  --lr=0.001  \
  --img-model=0 \
  --persuasive-label-threshold=0.5 \
  --save-checkpoint=0 \
  --skip-non-persuasion-mode=1

• Run text modality experiments

python main_text.py \
  --num-epochs=3  \
  --batch-size=16  \
  --exp-mode=0 \
  --data-mode=0 \
  --lr=0.001  \
  --img-model=0 \
  --persuasive-label-threshold=0.5 \
  --save-checkpoint=0 \
  --skip-non-persuasion-mode=1

• Run multimodality experiments

python main_multimodality.py \
  --num-epochs=3  \
  --batch-size=16  \
  --exp-mode=0 \
  --data-mode=2 \
  --lr=0.001  \
  --img-model=0 \
  --persuasive-label-threshold=0.5 \
  --save-checkpoint=0 \
  --skip-non-persuasion-mode=1

Citation

If you make use of this code, please kindly cite our paper:

@inproceedings{liu-etal-2022-imagearg,
    title = "{I}mage{A}rg: A Multi-modal Tweet Dataset for Image Persuasiveness Mining",
    author = "Liu, Zhexiong  and
      Guo, Meiqi  and
      Dai, Yue  and
      Litman, Diane",
    booktitle = "Proceedings of the 9th Workshop on Argument Mining",
    month = oct,
    year = "2022",
    address = "Online and in Gyeongju, Republic of Korea",
    publisher = "International Conference on Computational Linguistics",
    url = "https://aclanthology.org/2022.argmining-1.1",
    pages = "1--18",
    abstract = "The growing interest in developing corpora of persuasive texts has promoted applications in automated systems, e.g., debating and essay scoring systems; however, there is little prior work mining image persuasiveness from an argumentative perspective. To expand persuasiveness mining into a multi-modal realm, we present a multi-modal dataset, ImageArg, consisting of annotations of image persuasiveness in tweets. The annotations are based on a persuasion taxonomy we developed to explore image functionalities and the means of persuasion. We benchmark image persuasiveness tasks on ImageArg using widely-used multi-modal learning methods. The experimental results show that our dataset offers a useful resource for this rich and challenging topic, and there is ample room for modeling improvement.",
}