jack960317 / tma

Multimodal Sentiment Analysis with Temporal Modality Attention(Interspeech2021)

TMA

Multimodal Sentiment Analysis with Temporal Modality Attention

First check that the requirements are satisfied:

• Python 3.6.5
• torch 1.4.0
• scikit-learn 0.23.2
• tqdm 4.54.0
• numpy 1.19.4
• visdom 0.1.8.9

The next step is to clone the repository:

git clone https://github.com/qianfan1996/TMA.git

put your datasets into data/CMU-MOSI, data/MMMO, data/YouTube directory

You can run the code with:

python test_mosi.py 

in the command line. The default is to train MFN with TMA module, and gives a CMU-MOSI test set MAE of 0.960, binary classification accuracy of 77.4%.

In addition, you can change command-line arguments: -m [mfn/tman1/marn/tman2] set model types; -t set train or test.