AMIFN

Thank you for visiting this repository. This project proposes Aspect-guided multi-view interactions and fusion network (AMIFN) [paper]https://doi.org/10.1016/j.neucom.2023.127222

Data

• Twitter datasets : the processed pkl files are in floder ./data/Sentiment_Analysis/twitter201x/ .

data/Sentiment_Analysis/twitter2015: Text data from the twitter2015 dataset
data/Sentiment_Analysis/twitter2017: Text data from the twitter2017 dataset
data/Twitter_image/twitter2015_images: Image data from the twitter2015 dataset
data/Twitter_image/twitter2017_images: Image data from the twitter2017 dataset

• Generate train.pkl dev.pkl test.pkl (look at the file DataProcessor)
• Generate train.graph dev.graph test.graph (look at the file DataProcessor)
• The original tweets, images and sentiment annotations can be downloaded from [https://drive.google.com/file/d/1PpvvncnQkgDNeBMKVgG2zFYuRhbL873g/view]
• Download the pre-trained ResNet-152 via this link (https://download.pytorch.org/models/resnet152-b121ed2d.pth),rename to resnet152.pth and put the pre-trained ResNet-152 model under the folder './resnet/resnet"
• Download roberta-base https://huggingface.co/roberta-base/tree/main

Requirement

• torch
• transformers
• tqdm

Code Usage

Note that you should use your own data path. data_dir imagefeat_dir

bash train.sh

Acknowledgements

• Most of the codes are based on the codes provided by huggingface.
• Some code is based on the code of ITM many thanks! ITM