This repo is for the paper "Automatic-Controllable-Product-Copywriting-for-E-Commerce", which has been published in KDD 2022 ADS.
The folder named "stylized_product_copywriting" provides the whole process for building the generation model, inluding data preprocess, topic modeling, and generation model (unilm), as well as the postprocess. Follow the run.sh file to run the script.
The folder named "pretraininglm_ecommerce" is for pre-train the language generation model of uniLM, which will be finetuned for product copywriting generation.