Optimizing Search Engine Advertising Efficiency via Semantic Query Negation (Extended Abstract)

Abstract

This research introduces a proactive analytical framework that leverages natural language processing and semantic word embeddings to identify and negate non-performing keywords in Search Engine Advertising before significant costs accu-mulate. By utilizing InferSent to cluster queries based on se-mantic similarity, the proposed methodology offers a data-driven alternative to reactive industry heuristics, achieving campaign-wide savings of approximately 10% without any loss in revenue.