I'm not sure if I fully understand the explanations given here for how related questions are selected, but it seems like the basic idea is that it is based on similarity of words in the title, body, and tags of the posts.
My question is, does the system only look at at similarities or does it also take into account dissimilarities? For instance if Post A and Post B have 100 (non-trivial) words in common, and Post A and Post C have 101 (non-trivial) words in common, would A and C be more related than A and B even if A and B were exact duplicates while A and C also had 1,000 words not in common?
If this is indeed how it works, then the longer a post is the greater chance it has of being related to other posts simply by virtue of having more words. Taken to the extreme, an infinitely long post — even consisting entirely of nonsense/gibberish — would automatically be related to every other post.
The reason I ask, is that I noticed that Firing mods and forced relicensing: is Stack Exchange still interested in cooperating with the community? is related to Site for asking about a specific website even though on the face of it the two questions have nothing in common:
This led me to wonder whether the length of the first post (and the brevity of the second one) artificially inflated the relatedness.
