Abstract
Native advertisements mimic the look and feel of a publisher’s content slots and are used to monetize their inventories. Currently, native ad styles are created manually by the publisher based on hand-written rules and heuristics. This can result in ad styles that do not consistently resemble the look-and-feel of the publisher’s pages or apps. Also, current techniques generally use the DOM structure or HTML source of the publisher’s page or app to generate the native ad. However, the DOM structure or HTML source is not always available, e.g., for apps.
This disclosure describes the use of machine learning techniques to automatically generate native ad styles from key visual attributes of images. The images can be screenshots or design mockups. The techniques can generate native ad styles that match the publisher’s look-and-feel closely without recourse to the DOM structure or HTML source for the publisher’s page or app.
Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 License.
Recommended Citation
Lin, Bo; Pochampally, Ravali; and Yang, Yunfan, "Automatic generation of native ad styles using visual attributes of images", Technical Disclosure Commons, (May 06, 2019)
https://www.tdcommons.org/dpubs_series/2175