Most natural way to write sentences of the type "A verbs that B verbs that C verbs that..."

Question

A)

I think the typical way to say

Tanaka said that Yamata knows where the building is

in Japanese is

田中さんは山田さんが建物はどこか知っていると言った

Maybe it's more natural to say "the building's location" instead of using a interrogative clause, but the point is, the overall sentence structure takes the pattern

...S₃(S₂(S₁OV₁)V₂)V₃...

For small sentences, it's easy to keep track who is doing what. But when the recursion depth is larger, it gets harder to understand. Is there a way to reorganize the sentence such that the subject of each relative clause is next to the verb? I was thinking that it might be possible to make the sentences OSV order. Thus the pattern could be

...((OS₁V₁)S₂V₂)S₃V₃...

Here is my attempt, please tell me of any particle errors, or if the idea itself is flawed.

建物はどこか山田さんが知っているのは田中さんは言った

B)

The inspiration here is the Eddington controversy:

If A, B, C, D each speaks the truth 1 in 3 times (independently), and A affirms that B denies that C declares that D is a liar, what’s the (conditional) probability that D was speaking the truth?

in particular I'm trying to translate the bolded part

EDIT:

From the answers I realized I swapped 山田さん　and 田中さん

Answer 1

田中さんは山田さんが建物はどこか知っていると言った

To understand the word order in a complicated sentence like this, it is useful to break it down into parts called bunsetsu:

Each arrow represents that everything before the arrow modifies the bunsetsu immediately after the arrow. For example, 建物は modifies どこか, and both 山田さんが and 建物はどこか modify 知っていると.

(Two caveats:

1. It is probably more correct to say that 山田さんが and 建物はどこか modify 知っている, and that と is attached to the whole 山田さんが建物はどこか知っている, rather than 山田さんが and 建物はどこか modify 知っていると. But I am restricting the scope to the structure among bunsetsu, ignoring the structure inside them.

2. Not every sentence fits this model cleanly. There are sentences whose structure is explained better if we allow a modified part to consist of more than one bunsetsu. But I stick to the basic case here.)

A modifier must come somewhere before the bunsetsu it modifies (although there are exceptions (shameless plug)). Where a bunsetsu has more than one modifier, their relative order is flexible. So 山田さんが建物はどこか知っていると…… and 建物はどこか山田さんが知っていると…… are both grammatical. However, 建物は山田さんがどこか知っていると…… is not, because some part of the modifier 建物はどこか comes before another modifier 山田さんが and the other part comes after it.

(If you know the terminology in computer science, the diagram like the one above is a rooted tree where an arrow points from a child to a parent, and I have just claimed that any post-order traversal gives a grammatical ordering of words.)

Then there are some desirable properties of the word order:

• A subject should come first.
• A modifier and the bunsetsu it modifies should be close to each other, so a longer modifier is better to come earlier.
• Ambiguity should be avoided.
• Too many toten (、) should be avoided.
• …

In this case, I would give up putting the subject first, and write

建物はどこか山田さんが知っていると田中さんは言った。

(As you stated, it is simpler to use 建物の場所 in this sentence, but I am focusing on the order of the words.)

Returning to your attempt:

建物はどこか田中さんが知っているのは山田さんは言った

It is incorrect for several reasons. I hope that you can figure out why from the explanation above.

Answer 2

The very basic of Japanese word order is

1. Verbs come last.
2. Modifiers (including subjects) should be close to modified words (including verbs).
3. Clauses come first, phrases come second.
4. Longer modifiers come first, shoter modifiers come second.
5. If you want to invert them, use commas.

Of course, there are exceptions. For example, topics tend to come first with a comma.

---

A)

Tanaka said that Yamata knows where the building is

"その建物がどこにあるかは山田さんが知っていると、田中さんは言った", but more natural way in this case is "田中さんは「その建物の場所は山田さんが知ってるよ」と言った" or something like that.