Overview:
Investigate AI technologies for ways to help advertisers better optimize advertising dollars and systems.
Project Objective:
The goal is to create an easy-to-use system to generate ads using some user training and some AI. The idea is to create an interface where the end-user enters a URL from their site and generates ads rejecting ads that are not acceptable. The system will learn from this interaction and generate a custom tuned algorithm that can then be automatically applied to an entire website without any further user interactions. Ads can be generated automatically for a site including generating new ads for new pages as they are created by other processes.
Current Algorithms use a mix of existing useful phrases on the page, identifying valuable keyphrases to compile with templates, and the use of Markov Chains for natural language generation based on text on the page.
Caveats:
This system will not work with every website. Many sites use technologies and architectures that don't allow easy scraping and identification of useful data. Also, some algos work better with some site and not at all with others. Your selection of rejected ads will weed out non-useful algos and tune those that do work.
The Future:
We will continue to develop algorithms based on artificial intelligence and machine learning. We are also investigating ways to use non-text assets such as pictures.