Sentence Comprehension: Parsing
  1. Overview of comprehension. Comprehension involves mental processes that operate on language to yield understanding. Comprehension begins with the word, includes word meaning in context, sentences, and larger runs of text. It entails levels of language that include word meanings, morphology, and syntax.

  2.  
  3. The Parsing Problem: Assuming that one knows the meanings of words, sentence comprehension begins with parsing, the processes by which a linear string of morphemes (or words) is structured into meaning-related phrases.
    1. Illustrations of parsing with spoken language. The normal process is very rapid; you can observe some of what is involved by slowing down speech.
    2. Illustrations with written language
    3. The problem: How can a comprehender make rapid decisions about what goes with what-i.e. what the phrases are-when speech input is so rapid.
    4. The concept of a parser: A language mechanism that solves the parsing problem
      1. Decision points at each word/morpheme in a sentence. The alternative is to consider all possible (syntactically legal) continuations of the sentence.
      2. Garden Path Phenomena

      3.  
  4. Parsing Principles and Attachment Preferences
    1. Minimal Attachment: Build the simplest structure consistent with the structure currently built. Avoid complex attachments.
    2. Late Closure
    3. Others--Kimbal's principles

    4.  
  5. Examples of parsing problems and how these principles apply.
    1. While Mary was sewing the sock fell off her lap.
Constituents: Mary was sewing/ Mary was sewing the sock

Late closure: don't close a const. if you don't have to:

This produces a GARDEN PATH

    1. The spy saw the cop with the binoculars/revolver
Minimal attachment: don't build an embedded NP
    1. John heard the music would be good.
Minimal Attachment
    1. These examples are not too difficult to comprehend. The parsing problem is mild and there is a relatively easy recovery process. But the examples below illustrate more difficult problems. These are known as "reduced relatives" because they contain a relative clause ("that was raced", "that was spilled") that has been reduced by the deletion of the "that".
      1. The horse raced past the barn fell.
      2. The coffee spilled on the rug was difficult to conceal.
    2. These examples illustrate issues of configurational representation-how a reader or listener configures the input into phrasal units. This is essentially a question of using one's syntactic knowledge in combination with parsing principles

    3.  
  1. Lexical Information in Parsing. Configuration of phrases also depends on words
    1. Syntactic Category Information
    2. Verb Semantics and Argument Structure
      1. Arguments and Thematic Roles
        1. Verbs take arguments (e.g. nouns)
        2. The arguments have roles in meaning; agent; recipient; theme; experiencer, source, goal, location, etc.
      2. Put vs drop
        1. DROP (2 arguments): Agent, theme
        2. PUT (3 arguments): Agent, theme, goal (end location)
        3. Throw (agent, theme,[goal]
        4. Put (agent, theme, goal)

        5.  
  2. Related parsing issues: Complexity in syntax. Certain kinds of syntactic structures create problems not for a parser but for a limited capacity memory system:
    1. The cat chased the rat that ate the cheese.
    2. The cat chased the rat that ate the cheese that ...
    3. The rat the cat chased ate the cheese.
    4. The rat the cat the farmer bought chased ate the cheese.
    5. How long can this embedding go on? After "farmer" insert the woman loved bought chased ate...

    6.