Algorithmic Copywriting: Automated Generation of Health-Related\n Advertisements to Improve their Performance

Le résumé fourni par la source

Search advertising, a popular method for online marketing, has been employed\nto improve health by eliciting positive behavioral change. However, writing\neffective advertisements requires expertise and experimentation, which may not\nbe available to health authorities wishing to elicit such changes, especially\nwhen dealing with public health crises such as epidemic outbreaks.\n Here we develop a framework, comprised of two neural networks models, that\nautomatically generate ads. First, it employs a generator model, which create\nads from web pages. It then employs a translation model, which transcribes ads\nto improve performance.\n We trained the networks using 114K health-related ads shown on Microsoft\nAdvertising. We measure ads performance using the click-through rates (CTR).\n Our experiments show that the generated advertisements received approximately\nthe same CTR as human-authored ads. The marginal contribution of the generator\nmodel was, on average, 28\\% lower than that of human-authored ads, while the\ntranslator model received, on average, 32\\% more clicks than human-authored\nads. Our analysis shows that the translator model produces ads reflecting\nhigher values of psychological attributes associated with a user action,\nincluding higher valance and arousal, and more calls-to-actions. In contrast,\nlevels of these attributes in ads produced by the generator model are similar\nto those of human-authored ads.\n Our results demonstrate the ability to automatically generate useful\nadvertisements for the health domain. We believe that our work offers health\nauthorities an improved ability to nudge people towards healthier behaviors\nwhile saving the time and cost needed to build effective advertising campaigns.\n

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.