Codes for TriSAT: Trimodal Representation Learning for Multimodal Sentiment Analysis （Accepted by IEEE/ACM Transactions on Audio, Speech and Language Processing）.

Usage

Clone the repository

git clone https://github.com/gw-zhong/TriSAT.git

Download the datasets and BERT models

• CMU-MOSI & CMU-MOSEI (Glove) [align & unaligned] (which are not available now)

• CMU-MOSI & CMU-MOSEI (BERT) [align & unaligned]

Alternatively, you can download these datasets from:

• BaiduYun Disk code: zpqk

For convenience, we also provide the BERT pre-training model that we fine-tuned with:

• BaiduYun Disk code: e7mw

Preparation

Create (empty) folders for data, results, and models:

cd TriSAT
mkdir input results models

and put the downloaded data in 'input/'.

Run the code

python main_[DatasetName].py [--FLAGS]

Citation

Please cite our paper if you find that useful for your research:

@article{huan2024trisat,
  title={TriSAT: Trimodal Representation Learning for Multimodal Sentiment Analysis},
  author={Huan, Ruohong and Zhong, Guowei and Chen, Peng and Liang, Ronghua},
  journal={IEEE/ACM Transactions on Audio, Speech, and Language Processing},
  year={2024},
  publisher={IEEE}
}

Contact

If you have any question, feel free to contact me through [email protected].