This repository is the official implementation of the WSDM 2021 Paper Abstractive Opinion Tagging.

If you have any question, please open an issue or contact [email protected].

Please visit this URL to read the details of the eComTag dataset.

Please use this Google form to submit your information and request access to eComTag.

Check the packages needed or simply run the command:

pip3 install -r requirements.txt

For reproducibility purposes, we place the model checkpoints at Google Drive. You could download and move it under /output/.

We would now set up our directories like this:

.
└── model
    └── ...
└── baselines
    └── ...
└── eComTag
    └── ...
└── utils
    └── ...
└── README.md

Training models associated with AOT-Net will take more than one day. You can also run the trained models to test by changing "False" to "True" in the bash commands.

AOT-Net

bash script/AOTNet.sh 0 False

w/o SSE

bash script/woSSE.sh 0 False

w/o RCR

bash script/woRCR.sh 0 False

w/o AF

bash script/woAF.sh 0 False

w/o AL

bash script/woAL.sh 0 False

RNN

bash script/RNN.sh 0 False

PG-Net

bash script/PGNet.sh 0 False

Transformer

bash script/Transformer.sh 0 False

If you find our code useful, please cite our paper as follows:

@inproceedings{li-etal-2020-aot,
  title={Abstractive Opinion Tagging},
  author={Qintong Li and Piji Li and Xinyi Li and Zhaochun Ren and Zhumin Chen and Maarten de Rijke},
  booktitle={WSDM},
  year={2021},
}