Identification of negative keywords in search marketing with embedding layers and neural networks

Abstract

[en] In this paper we introduce a model to help marketing specialists within the field of Search Advertising to limit spend on Google searches which have a low probability of leading to a revenue generating event. This is a topic which has not been widely addressed in scientific literature. For this study, we obtained data from a company which spends a large amount on Google Ads, but relies on a subjective and time-consuming approach to this problem. Our proposed model uses GloVe’s pre-trained Embedding Layers and Neural Networks to speed up and improve accuracy of this process.