MMIB

Code for the the paper "Enhancing Multimodal Entity and Relation Extraction with Variational Information Bottleneck", which is accepted by TASLP2023

Model Architecture

Data

We followed Chen et al.2023 and Zhang et al.2023 for data processing.

Commands to run the code:

CUDA_VISIBLE_DEVICES=1 python main.py --dataset_name twitter15 --num_epochs=30 --batch_size=8 --lr=3e-5 --warmup_ratio=0.03 --eval_begin_epoch=1 --seed=0 --ignore_idx=0 --max_seq=128 --log_name twitter15_model --do_train

CUDA_VISIBLE_DEVICES=1 python main.py --dataset_name twitter17 --num_epochs=30 --batch_size=8 --lr=3e-5 --warmup_ratio=0.03 --eval_begin_epoch=1 --seed=0 --ignore_idx=0 --max_seq=128 --log_name twitter17_model --do_train

CUDA_VISIBLE_DEVICES=1 python main.py --dataset_name MRE --num_epochs=30 --batch_size=8 --lr=3e-5 --warmup_ratio=0.03 --eval_begin_epoch=1 --seed=0 --ignore_idx=0 --max_seq=80 --log_name MRE_model --do_train

Citation

If our paper is helpful for you, please cite as (The TASLP citation is coming soon):

@article{DBLP:journals/corr/abs-2304-02328,
  author       = {Shiyao Cui and
                  Jiangxia Cao and
                  Xin Cong and
                  Jiawei Sheng and
                  Quangang Li and
                  Tingwen Liu and
                  Jinqiao Shi},
  title        = {Enhancing Multimodal Entity and Relation Extraction with Variational
                  Information Bottleneck},
  journal      = {CoRR},
  volume       = {abs/2304.02328},
  year         = {2023},
  url          = {https://doi.org/10.48550/arXiv.2304.02328},
  doi          = {10.48550/ARXIV.2304.02328},
  eprinttype    = {arXiv},
  eprint       = {2304.02328},
  timestamp    = {Mon, 17 Apr 2023 15:20:10 +0200},
  biburl       = {https://dblp.org/rec/journals/corr/abs-2304-02328.bib},
  bibsource    = {dblp computer science bibliography, https://dblp.org}
}