This research paper addresses the need for a benchmark in automatic ad text generation (ATG) due to the limitations of manual online ad production. The authors propose a benchmark dataset, CAMERA, for ATG, designed to leverage multi-modal information and conduct an industry-wide evaluation. The paper also provides evaluation experiments using multiple baseline models and discusses the current state of ATG and future challenges. The discussion includes the use of pre-trained language models and the incorporation of multi-modal information, highlighting the need for a shared dataset that can be universally applied across the field.
Publication date: 21 Sep 2023
Project Page: https://arxiv.org/abs/2309.12030v1
Paper: https://arxiv.org/pdf/2309.12030
