About the meaning. Syntax subordination trees (subordination trees)

    In a previous post, I talked about subordination trees and used (perhaps in vain) a controversial example of the "nebula".
    Now it’s just necessary to explain why I interpreted this text in this way. Although, as it turned out, according to the comments of experts in Pushkin’s work, it’s incorrect, but we will consider this example not in terms of historical accuracy, but in terms of ways of interpreting text in a natural language by machine.
    Let's start with the definition of what are the trees of syntactic subordination (in common space - subordination trees)? This is an ordered graph (i.e. a tree), where the nodes are the words of the sentence, and their hierarchy and subordination system determines which words are main in the sentence and which depend on which.
    For clarity, I’ll give a couple of shots of what I mean:

    Presentation option number 1:
    Presentation option number 2:

    I will give all other examples in the second option (although I myself think the first is more convenient), but since most of the material is in the second - then do not exact.

    The main word in such a scheme is always the predicate - so that each sentence is considered as a process of something (birth, gift, arrival, etc.), from which all additional information is then extracted. Just as from triplets of ontologies begin to clutch at the middle (a bunch of subjects and objects).

    The main criterion for evaluating the correctness of construction and _interpretation of a sentence in such a scheme is the projectivity property of the subordination tree. It’s simplistic to say that a subordination tree is projective when its branches do not intersect.

    Not projective:

    This, of course, is a simplified model - there are mathematically described degrees of projectivity (strictly projective and weakly projective trees) that define different restrictions on the structure of the sentence so that it can be considered projective.
    The model of projective trees has little application to fiction and poetry, where the author, trying to express his feelings or give the sentence more expression, violates many rules of syntax and the accepted word order.
    Although in ordinary texts, the lack of projectivity is a sign of the lack of literacy of its compiler.
    Projective trees are rather rare in fiction.
    Well, now the same controversial example:
    Note that the upper tree of subordination (with a standard meaning) is just projective.

    The model of subordination trees is as old as the world (at least 60 years) - it has a bunch of comments, a bunch of caveats and is not a model applicable for all cases - there are more modern (and more complex models): for example, systems of components. But thanks to this model, you can get a fairly large number of characteristics of the proposal itself.

    In many modern semantic analysis algorithms, this model is not used in its pure form (and the projectivity rules are rejected by many and replaced by others), but the basic ideas are still alive.

    Also popular now: