You can not select more than 25 topics
Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
51 lines
2.5 KiB
51 lines
2.5 KiB
From text to speech: The MITalk system
|
|
|
|
vocalic suffix. The “optional” category is used for all other morphs which permit
|
|
spelling changes. Examples are:
|
|
|
|
e required - scar+ed —scarred
|
|
|
|
o forbidden - alloy+ing — alloying
|
|
|
|
e optional - change+able — changeable, change+ing — changing
|
|
|
|
The spelling changes performed by DECOMP consist of appending a letter,
|
|
or changing or deleting the last letter. When a morph is matched to the left end of
|
|
a word, the following procedure is used to determine the set of possible spelling
|
|
changes: If the matched morph is not a suffix, then no spelling changes are made
|
|
and recursion proceeds normally. If the matched morph is a nonvocalic suffix and
|
|
the last remaining letter is an i, then the change i —>y is examined. If the matched
|
|
morph is a vocalic suffix, then spelling changes are performed by matching a
|
|
three-character template against the last two remainder letters and the first letter of
|
|
the matched morph.
|
|
|
|
If the vocalic suffix is es, then a special check is made to determine whether
|
|
the letter e should be considered part of the suffix. If the remainder does not end
|
|
in ¢, ch, g, i, o, s, sh, x, or z, then the es match is immediately rejected. This
|
|
causes the morph s to be the next match; the e is thus moved from the suffix to the
|
|
remainder. This rule is motivated by the phonetic properties of the plural suffixes
|
|
s and es. The lexicon entry for s gives a pronunciation zz while the entry for es
|
|
gives 18 2zz. The rule allows DECOMP to make the appropriate decomposition
|
|
of tunes— tune+s rather than tunes— tune+es which is found first. The two
|
|
vowels, i and o, are permitted to precede es to enable correct decomposition of
|
|
words such as heroes and parties, even though pronunciation is not correct in such
|
|
cases; morphophonemic rules are used in a later MITalk module to obtain the
|
|
proper pronunciation. |
|
|
|
|
Table 3-1 shows the set of template patterns and their resulting spelling
|
|
change actions. The plus sign (+) in the pattern denotes the boundary between the
|
|
suffix and the remainder. A dot (.) in the pattern matches any letter. The pattern
|
|
xx matches any doubled letter. The first pattern (from top to bottom of the list)
|
|
which matches the current remainder/suffix pair controls the set of changes ap-
|
|
plied.
|
|
|
|
For each possible spelling of the remainder, the following steps are per-
|
|
formed:
|
|
|
|
1. Make the change.
|
|
|
|
2. Recursively decompose the remainder.
|
|
3. If a morph matches the right end of the remainder, check its spelling
|
|
|
|
34
|