AdvertiseGen

Dataset introductionAdvertiseGen is an e-commerce advertising copy generation data set.
AdvertiseGen is based on the information correspondence between product webpage tags and copywriting. It is a typical open generation task. When the model generates open copywriting based on key-value input, the factual consistency with the input information needs to be focused.
Task description: Given the keywords and attribute list kv-list of product information, generate advertising copy adv suitable for the product;
Data size: training set 114k, validation set 1k, test set 3k;
Data source: Tsinghua University CoAI Group;