For the following I want to note that I am german. So if you recommend any service with a given database, it has to work with german phrases(aswell).

As for your question: I'm not aware of software to do that. I think people at places
like Google might be working on stuff like that, or at least I imagine they
ought to, or at least I'd like them to.
Nice.

Google brought the prediction api to us, which is a kind of uclassify. But you cannot test the pa without billing and uclassify, well, might work. But for privacy reasons I prefer a local version. After all it is quite private what I am collecting there obviously.

Software could work like this: Take a very large corpus such as a collection of a large
number of web pages. Rate words or phrases as related if they appear more often near each other (e.g. within a couple of paragraphs) than randomly. Given two or more words or phrases, rate other words or phrases as interesting if they're related to more than one of the other words or phrases. A score for how interesting could be calculated. Print out the most interesting ones.
Yes. Do you know a general term for this so I could search for appropriate c# snippets? I would do it the way you suggested, this was my original idea. But programmatically how would you go about it?
First separate all words by space and/or punctuation. This leads to an array. Then taking the first and second entry, saving them somewhere. Then looking through the whole array to find the same two words in this order. Then assign a number of occurences to it in a new associative array. Then a synonyms-dictionary has to come into play. What was our goal again? I feel we are losing track here.

So, my approach (which need not be the same as your approach) would be
to look over the material regularly as Ternouth suggests, perhaps varying it
somehow -- shuffling the order, rewriting it in different words, etc. --
and wait for ideas to come bubbling up out of my subconscious.
Writing down your dreams can help, too. And solving cryptic crosswords.
Which question are you answering here?

I hope you find a computer program that does what you're talking about.
Let me know if you do, because I'm interested in that kind of thing.
I appreciate your interest. As mentioned above this is the route I want to go to.

PS:
1. Can you elaborate the connection between my task and Hypothes.is?
2. You seem competent, can you exert your wisdom in your day to day job?