I fear it's incredibly hard to fight fake content with automated systems. You need the ability to understand the content to determine whether the content is actually meaningful. For that, you need far better intelligence than just some fancy text processing.

You need people, basically. If you want to automate it, you need some way to figure out what people think about that content. This has been tried by counting how many people link to it, or rate it highly, but those methods have also been gamed to death.