at some stage, we will need to convert the phoneme strings under "IPA"
into phonological structures of the kinds that I sent a grammar for the
other day. I think we should talk about how this will be done.
Would it be possible to keep our structured representations alongside
lexical entries, so that once something is parsed it needn't be parsed
again, but just looked up?
A related question, more to do with Alex's "small lexicons" point: how
feasible would it be to have a lexicon of all things that behave in
word-like ways? I'm thinking in particular of Germanic affixes like
<-less> and <-ness>, which seem to join to other morphemes in the same
sort of way that proper words do; and in Mark's productions that I
listened to this morning, they have different vowels in them (quite
[E]-like in "less" and quite [I] like in "ness", though both are
transcribed as [@] in the lexicon).
A couple of things I don't understand in the lexicon:
"the" is ADV(ge). what does this mean?
"was", "were" (etc.) have full forms in the lexicon, no weak forms(*). how
are we going to get weak forms out? there's no labelling of eg. "was" as
an auxiliary, which I think may be needed. These words are so recurrent in
the material, that I think we need to consider carefully how they get
represented. what is more they are, in our material, regularly metrically
weak, not strong, so that the full vowel forms don't occur. (But I admit
to also having a bee in my bonnet on this one -- though the function words
are so ubiquitous that we've got to get them right.)
(*) Though I note that "to" has both [t@] and [tu], "an" has [@n] and
[&n].
the transcriptions imply that words like "paddle" have only one syllable
because they are done as p/A/dl, where // surrounds the nucleus. I don't
understand how consequential this might be, but perhaps we need to have it
as p/&/d/@/l or p/&/d/l/ so that it's clear there are two syllables.
(Clearly it's not the same shape as a word like r/I/ns.)
sorry for the pulic display of obsessive behaviour!
Richard