Don't Recommend Constituent Works

I've just taken a look at fiction books recommended for me. Besides the high number of books that shouldn't be recommended here anyway (I've complained about that here), I noticed that constituent works get recommended even though I have them all in bundled form. Mostly I see this with J.R.R. Tolkien's Lord of the Rings trilogy. I have this one-volume edition, but still got recommended The Two Towers under popular books I don't have. Problem is, I do have it - but it counts as a different work because it's bundled with The Fellowship of the Ring and The Return of the King.

Can this be fixed?

LT doesn't "know" that The Two Towers is "inside" The Lord of the Rings (too many movies and other weirdnesses to touchstone those right). We will need a super-/subwork concept for this to work properly.

Would that be particularly difficult to do? I'll admit, it's only one aspect of a larger offness in the recommendations, but it's still an issue.

Well, it's been a suggested improvement pretty much since works have existed as a concept, and I've seen Tim say in the past that he has some misgivings about how difficult it would be to implement.

3: Yes. :)

Database design/structure, stability, content (who decides what works contain other works, how far does it go down, etc).

AnnaClaire @3, it would require adding another layer entirely to the structure -- actually, potentially several, since _The Two Towers_ is contained by _The Lord of the Rings_, but it contains volumes 3 and 4 of the seven-volume edition (which someone concievably could have entered as separate volumes.) Then there are the many omnibus editions combined in different ways, where you can have Book X by an author existing on its own, and in editions combined with Book Y or Book Z (separately), or with both of them....

And then there are the people entering short stories -- they may want to start associating a given anthology with each story in it, and a given story with each anthology in which it appears. It gets ugly fast.

Perhaps an interim solution to this (and other) recommendations issues would be a "not interested" button by each of the recommendations?

That way, the LT database doesn't have to be structured to "know" anything about the meta/sub-works, and people will only be annoyed by the "I already have this"-type of recommendations once.

That might work, but it's only a good interim solution for this problem if the removing the sub-work in question from what you've been recommended doesn't prejudice your recommendations against books like the one you're removing as well.

7: That would be good too, but for entirely different reasons. :)

6: Do you really need add another layer to the structure? Surely any work is still a work, but might have pointers to multiple children and/or multiple parents...

If ∈ (is an element of) means "is a child of", then

Book III ∈ The Two Towers ∈ The Lord of the Rings

and

Book III ∈ The Lord of the Rings

are both true/possible relationships.

But I guess that would make listing the children of LOTR ... odd ...