Modeling human reasoning 7 Way. Search the way.

(LION) -> [EAT] ->? -> (CAT) Is a lion eating a cat?
That is our task to check:
1) The presence of such a connection in the database.
2) Make sure the connection is positive.
3) Give the result as a response (Yes / No)
4. If there is no such connection, then build such a connection (if possible)
We climb into the database and see that the connections are:
(LION) -> [EAT] -> YES -> (CAT) we do not have.
(LION) -> [EAT] -> NO -> (CAT), we also do not.
So if there is no negative connection (dead end, end of the way), then the connection can be positive.
And if so, then we must look for a way.

And then we are looking for a match based on the data we have.
> What does "lion" mean?
We look at the "description" (the concept of what) and understanding.
UNDERSTANDING is a “description” but written in an internal language. It’s difficult to handle the "description"
This is text info, it is not a thought. We need a thought. A thought is a connection.
UNDERSTANDING is the very type of connection (LEV) -> [ЕСТ] -> YES -> (CAT) obtained from words and stored in the database.
Called: "bunch."
Not confused? Then further:

Build a path. You rightly said, we start with the process, as he is in charge here.
With the word lion can be dofiga and more links. But "eating" tied to the "Leo" will truncate our "bunch" to a minimum. That is, to a possible solution of the problem.
And so: "eats".
(LION) -> [EAT] ->? -> (CAT) Is a lion eating a cat?
unknown
(LION) -> [EAT] -> YES -> (?)
What does a lion eat?
We look at the database. If there is such a connection (branching)
For example:
(LION) -> [EAT] -> YES -> (ANIMALS)
it means we already have a “path”: the lion eats animals. Check by substitution:
(X) -> [EAT] -> YES -> (ANIMAL) Replace with a variable (i.e. lion = X = animal)
Does the variable match the condition?
(CAT) -> [IT] ->? -> (ANIMAL)
If there is such a record (branching)
(CAT) -> [IT] -> YES -> (ANIMAL)
Then we substitute instead of X (LEO) -> [EAT] -> YES -> (CAT). The problem is solved.
If there is no such record: The bad thing is, we don’t know whether the animal is a cat.
WHAT do we have a cat? We look at the connection one by one and check.
Suppose: a mammal cat. Lion eat mammals? Etc.
Nifiga found nothing. Then we go from the other end:
(?) -> [EAT] -> YES -> (CAT)
Who eats cats?
Suppose there is a record:
(DOG) -> [EAT] -> YES -> (CAT)
change to X (i.e. dog = X = lion)
(X) -> [EAT] -> YES -> (CAT)
Check the substitution:
(LION) -> [IT] ->? -> (DOG)
Is a lion a dog? The answer is no. (the concepts are WHAT)
Poorly. The task is not solved.
We look who STILL eats cats (alternately)
/ this is called deep into /
Not a fig found. Poorly.
Look again, "eat." How to eat? That is, we take out the description of the process.
> What does "eat" mean?
Now we are building a relationship satisfying the condition.
I.e:
(?) -> [EAT] -> YES -> (?) Who eats whom?
For example:
(ANIMAL) -> [EAT] -> YES -> (ANIMAL)
Similarly, replacing with X and Y and checking whether LEO and CAT are the desired X and Y
If not now found.
/ we leave aside /
We look "process description"
> the concept of "eat" - this is when the digestive process and all that ...
and further similar to the previous example:

[EAT] -> [IT] -> YES -> [?]
or so [ЕСТ] = [?] search for synonyms (with the condition, for example, "the presence of the digestive tract" or without, for example EAT)
(?) -> [X] -> YES -> (?)
And by substituting instead of X, for example: (LEV) -> [EAT] -> YES -> (CAT) are looking for matches.
Let's say a dead end.
Then the last option. We build a temporary relationship:
(LION) -> [EAT] -> ALLOW -> (CAT)
and start checking:
Could lion eat cats? Once we have reached here, it means that we don’t have a direct recording, and there is no point in checking again. But we have "connections through connections" that we haven’t yet verified ...
Further while it makes no sense to write.

Question: Is it clear at least something?