Re: PIE Soundchanges - Grassman & Bartholomae

Henrik Theiling skrev:

> This is because the replacement is allowed to contain group names if
> these are used in the match as well.  The compiler then replaces the
> group name in the replacement by the internal $<number> it uses for
> that group in the match.  This only works when the group name is
> unique in the match, which is why you cannot refer to C in a match
> like ChC (there are two C's).  The above rule can be rewritten to move
> one C into a context, but this need not be possible for arbitrary
> metathesis rules.  Extending this is on by TODO list.
By trial and error (and peeking in the SCHcompiler output module)
I've found that

     (C) \3  >  C

works for degemination and

     C  >  _ _

works for gemination.  This probably is exploiting a bug
but it saves one from having to write a subrule for each
consonant.  I had to write a subrule for each consonant
in order to capture Sohlob palatalization, although a
text editor with regexp search-and-replace helped a lot...
(Obviously you could write all geminates as _t:_ etc.
in your input, but you may want to squeeze geminates
that arise from syncope for example.  The same goes
for sequences of like vowels that arise through
consonant loss or umlaut or provection that you may
want to convert into vowel + length mark...

>>More  to the point, given the syllable-based approach of SCH, is a
>>rule like  that even easily to implement across syllable borders?
>
>
> Yes, you can match on several syllables at once by not putting
> adjacent syllable constraint into brackets: the string the replacement
> rules will match is a concatenation of the non-bracketed syllables.
> E.g.:
>
> rule "blah"
>    syllable [ first ], any, any
>    lV > Vl / k_V
>
> With the right input encoding, this would transform 'nuclear >
> nukelar' (not exactly what was intended, but you get the point).
This will apply the rule anywhere but in the first syllable?
Good to know.  FWIW writing

     lV  >  Vl / ^C*VC (C,V)* k _ V

should do the same.  I *had* to use this kind of notation
when the content or structure of adjacent syllables mattered,
as in the Romance stress rules.  Sievers' rule (not the same
as Sievers' Law!) became:

     rule "Siever's rule"
         -- g^w/G^w > g/G after the accent
         \^w > 0         / ' (C,V)*(g,G) _
         -- g^w/G^w > w before the accent
         (g,G)\^w > w    / _ (C,V)* '


--
/BP 8^)>
--
Benct Philip Jonsson -- melroch at melroch dot se

    "Maybe" is a strange word.  When mum or dad says it
    it means "yes", but when my big brothers say it it
    means "no"!

                            (Philip Jonsson jr, age 7)

Reply