Data structure

The main data structure of IPOX is the metrical tree, or, more generally, a binary-branching, headed tree.

Type definition:

   Tree   ::   Node --> Tree
           |   ( Tree / Tree )
           |   ( Tree \ Tree )
           |   Terminal

In this definition, the slash is used to indicate which of two sister nodes is the (prosodic) head.

Graphic representations:

conventional metrical tree: only weak-strong or strong-weak is indicated; unary nodes are ignored
headed constituent tree: constituent labels are shown, as well as headedness

Example:
two representations of the prosodic structure of the word photographical:

metrical tree headed tree

Internal representation:

   W --> 
      (F --> (S --> "pho" \ S --> "to") 
      / F --> ((S --> "gra" \ S --> "phi") \ S --> "cal"))

A node in a metrical representation is a complex feature structure (Shieber 1986).

Type definition:

   Node   ::  Category:Features

   Features :: [FV*]

   FV     ::  Node
           |  Feature=Value
           |  +Feature          (Feature=true)
           |  -Feature          (Feature=false)

Example:
complex feature structure for the unstressed vowel schwa

   vow:[-stressed, voc:[+back, -rnd, height=mid]]

Our data structure is not limited to the representation of prosodic headedness.

Other uses:

syntactic headedness:
```
   VP --> (V \ NP)
```
morphological structure:
```
   word --> ((prefix / root) \ suffix)
```

Objection:
Our data structure is too general because prosodic structure is "flat", that is, non-recursive (Nespor and Vogel 1986).

Defense:

self-embedding is OK if across the board , as in (a) below
self-embedding is compiled away in (b), obtaining (c)
a prosodic category may not dominate a category of a higher type

Examples:

(a) (b) (c)

stress patterns of two Dutch words: (a) universiteit 'university'; (b=>c) makaroni 'macaroni'

Back to Table of Contents

Arthur Dirksen / adirksen@prl.philips.nl / January 13, 1995