SUSTAINABLE SIGNALS: An AI Approach for Inferring Consumer Product Sustainability

This folder contains the code for the paper《SUSTAINABLE SIGNALS: An AI Approach for Inferring Consumer Product Sustainability》

Requirements

datasets==2.7.1

numpy==1.20.1

pandas==1.2.4

scikit_learn==1.0.2

scipy==1.6.2

torch==1.12.1

tqdm==4.59.0

transformers==4.25.1

We pre-trained several models. If you want to run the models other than ''Cate'', please download the folders ''distilroberta-envclaim'' and ''pre_review'', and put them in the directory ''deep_learning_package''. The pre-trained ''distilroberta-envclaim'' and ''pre_review'' models are open on request.

Usage

The main implementation of SUSTAINABLE SIGNALS is in the folder ''deep_learning_package''

The hyperparameters for the SUSTAINABLE SIGNALS can be found in ''option.py''.

If you want to run the model, please make sure the working directory is in ''deep_learning_package'' and use the command:

python3 main.py \
    --num_epochs 13 \
    --mode onTest \
    --train_batch_size 32 \
    --test_batch_size 32 \
    --model_type Cate \
    --model_name distilbert-base-uncased \
    --save_logging_steps 500 \
    --learning_rate 5e-5 \
    --house_dim 768 \
    --beauty_dim 768 \
    --baby_dim 512 \
    --kitchen_dim 512

You can also run

bash run.sh

to generate our results in the paper.

Data

The data in the repo are the synthetic data just for testing the code. We have provided text annotations for the reviews at this link: https://socialmediaarchive.org/record/45?&ln=en.

Output File

You can find the outputs of our model in the ''output'' folder, and testing results in the ''expe'' folder.

Citation

@inproceedings{sustainability_signals,
  title     = {{SUSTAINABLESIGNALS: An AI Approach for Inferring Consumer Product Sustainability}},
  author    = {Lin, Tong and Xu, Tianliang and Zac, Amit and Tomkins, Sabina},
  booktitle = {Proceedings of the Thirty-Second International Joint Conference on Artificial Intelligence},
  year      = {2023},
  note      = {AI and Social Good Track. The first two authors contributed equally.},
}