LINGUIST List 4.674

Thu 09 Sep 1993

FYI: Errata for "Morphology and Computation"

Editor for this issue: <>

Directory

1. Richard Sproat, Errata for "Morphology and Computation"

Kenneth Beesley kindly brought to my attention some technical errors
in my recent MIT Press book, "Morphology and Computation". Since I
have no idea if or when MIT Press would publish an errata sheet, I
decided that LINGUIST would be appropriate for this purpose.

1. (quoting more or less directly from Beesley's note)

Dental n to retroflex n (pages 134-136)

The description and formal statement of the rule (p. 134) state that
the mapping from lexical n to surface retroflex n occurs "when
preceded by a continuant retroflex consonant . . . somewhere in the
word, provided that there is no intervening palato-alveolar,
retroflex, or dental consonant and provided that a sonorant follows
the /n/ in question."

The table (p. 135 (9)) and the diagram (p. 136) contain a state 2,
which is the state where the left context has been found. From state
2, /n/ to retroflex n causes a transition to non-final state 3, from
whence a sonorant (the right context) is the only escape. That much is
OK. However, from state 2 these machine fail immediately on input of
n:n; and this failure is premature. From state 2, n:n should be legal
unless it is followed by a sonorant (the right context). So we need a
new state 4, and the transitions would look roughly like the following

 1: 2 2 1 0 1 1 1
 2: 2 2 1 3 4 2 1
 3. 0 1 0 3 0 1 0
 4: 2 0 1 0 4 0 1

This should replace (9) on page 135.

2. l->r intervocalically (page 138)

The FST compilation of the rule in (10) allows ara:ala but it fails to
block ara:ara. The correct FST is:

 r r V =
 l ===
1: 0 1 2 1
2: 3 4 2 1
3. 0 0 2 0
4: 0 1 0 1


Richard Sproat
Linguistics Research Department
AT&T Bell Laboratories | tel (908) 582-5296
600 Mountain Avenue, Room 2d-451 | fax (908) 582-7308
Murray Hill, NJ 07974, USA | rwsresearch.att.com