|
|
S
|
|
|
|
|
|
Morphophonemics and stress adjustment
|
|
|
|
|
|
5.1 Overview
|
|
|
|
|
|
It is not always possible to simply concatenate the pronunciations of the con-
|
|
|
stituent morphs of a word to get its pronunciation. There are sometimes changes in
|
|
|
pronunciation at morph boundaries. Module SOUNDI1 checks for contexts in
|
|
|
which such changes occur, and changes the pronunciation. It also adjusts the lex-
|
|
|
ical stress for compounds and for words having suffixes requiring special stress
|
|
|
rules. SOUNDI1 also performs letter-to-sound conversion for words which were
|
|
|
not segmented by DECOMP (this function will be described in the following
|
|
|
chapter). It accepts as input all the word and morph information given by
|
|
|
DECOMP and the additional phrase part-of-speech information produced by
|
|
|
PARSER. Output is a set of phonetic segment labels for each word along with the
|
|
|
phrase information from PARSER.
|
|
|
|
|
|
5.2 Input
|
|
|
Input to SOUNDI is the output stream from PARSER. The format of this stream
|
|
|
has been described in Chapter 4. It contains morph pronunciation information
|
|
|
|
|
|
from DECOMP and phrase and part-of-speech information from PARSER.
|
|
|
|
|
|
5.3 Output
|
|
|
|
|
|
The output stream from SOUNDI consists of a string of phonetic segment labels,
|
|
|
stress marks, and syllable and morph boundaries for each word. At the end of each
|
|
|
sentence, the phrase information for that sentence is placed in the output stream
|
|
|
(this is simply a duplicate of the phrase information from PARSER).
|
|
|
|
|
|
5.4 Morphophonemic rules
|
|
|
|
|
|
The pronunciation for each word which has been segmented by DECOMP is con-
|
|
|
structed by catenating the pronunciations of its component morphs. The following
|
|
|
rules are applied to modify the morph pronunciations when necessary.
|
|
|
|
|
|
5.4.1 Plurals, possessives, and contractions with “is”
|
|
|
|
|
|
Words which end in a fricative or affricate close in place of articulation to ss and
|
|
|
232, i.e., the set of segments Ss, 2z, SH, zH, CH, and JJ, form their plurals and
|
|
|
possessives by the concatenation of the segment string IB 22, or, in its vowel-
|
|
|
reduced form, Ix 2z2z (e.g. busses, churches, garages, marsh’s). After other
|
|
|
|
|
|
52
|
