Thanks for your reply! Indeed, I had thought of creating such rules manually. They all were very basic and would not have to be context-sensitive (as in your bed English example) Common superlatives get highlighted, and some silly, overused phrases too. On the other hand, we would like to whitelist multi-word constructions LT doesn’t know and highlights without good reason. I have raised this request before.
My impression is that extensions of the personal library functionality which only whitelist forbidden strings and greylist questionable wording would not require running LT on our server. Still, thanks for sharing the link – currently though, we have too many other fish to fry :o)
The only automated functionality I hinted was searching for term frequency in a defined body of text (a web-domain). Is is something one can achieve (with reasonable effort), using the self-hosted version of LT?