This kind of thing isn't anything new... here's a similar system that is less geared toward vandalism, but more to pointing out structural and lexical flaws in

This kind of thing isn't anything new... here's a similar system that is less geared toward vandalism, but more to pointing out structural and lexical flaws in the articles: https://github.com/jcla1/wikipedia_analyser https://github.com/jcla1/wikipedia_analyser (Disclaimer: I'm the author) It too uses neural-nets (and an up to 3-gram model) to detect flaws...