The source code and datasets for LREC-COLING 2024 paper: MMAPS: End-to-End Multi-Grained Multi-Modal Attribute-Aware Product Summarization in E-commerce [arXiv preprint].
models
folder contains the image encoderimg_transformer.py
, and the overall frameworkmodeling_bart.py
.utils
folder contains the data processing filedata_helper.py
, and the metric filemetric.py
.
The required environment is included in requirements.txt
.
The dataset used for experiments is a Chinese E-commerce Product summarization dataset CEPSUM
To train the model:
python main.py --mode train
To test the model:
python main.py --mode test