Researchers have developed a novel approach to enhance advertising models by integrating user onsite and offsite conversion data into a large-scale heterogeneous graph. This method utilizes a Knowledge Graph Embedding (KGE) model called TransRA, which was adapted to better incorporate graph embeddings into Ads ranking models. After initial challenges, an attention-based KGE finetuning approach and the Large ID Embedding Table technique were employed, leading to significant improvements in Click-Through Rate (CTR) and Conversion Rate (CVR) prediction. The framework has been successfully deployed on Pinterest's Ads Engagement Model, resulting in a notable CTR lift and CPC reduction. AI
IMPACT This research demonstrates a practical application of graph embeddings and KGE models to significantly improve advertising performance metrics.
RANK_REASON The cluster contains an academic paper detailing a new methodology for entity representation learning and its application in an industrial setting. [lever_c_demoted from research: ic=1 ai=1.0]
AI-generated summary · Google Gemini · from 1 sources. How we write summaries →
