Automatic Creative Selection with Cross-Modal Matching

TL;DR

The paper tackles matching app promo images to short search phrases by fine-tuning a cross-modal model (based on LXMERT) on in-house (search phrase, image, label) data. It introduces a mid-fusion cross-modal encoder that processes text and image modalities separately before combining them, achieving higher accuracy than CLIP-based and transformer+ResNet baselines. Results show notable AUC gains on both advertiser-associated and human-labeled ground truth, validating the approach's effectiveness in the search-ad domain. The work offers practical benefits for automatic image selection and self-serve image recommendation for developers, enabling more relevant creative pages for user queries.

Abstract

Application developers advertise their Apps by creating product pages with App images, and bidding on search terms. It is then crucial for App images to be highly relevant with the search terms. Solutions to this problem require an image-text matching model to predict the quality of the match between the chosen image and the search terms. In this work, we present a novel approach to matching an App image to search terms based on fine-tuning a pre-trained LXMERT model. We show that compared to the CLIP model and a baseline using a Transformer model for search terms, and a ResNet model for images, we significantly improve the matching accuracy. We evaluate our approach using two sets of labels: advertiser associated (image, search term) pairs for a given application, and human ratings for the relevance between (image, search term) pairs. Our approach achieves 0.96 AUC score for advertiser associated ground truth, outperforming the transformer+ResNet baseline and the fine-tuned CLIP model by 8% and 14%. For human labeled ground truth, our approach achieves 0.95 AUC score, outperforming the transformer+ResNet baseline and the fine-tuned CLIP model by 16% and 17%.