>I attach three component files for Phase1 DB:
>
>1. phase1.syn:
> fun = function
> cat = category
> att = attributes
> lex = lexical item
> time = temporal location
>
>2. phase1.wrd
> lex = lexical item
> time = temporal location
> ipa = IPA transcription
> stp = stress pattern (s=strong, w=weak)
>
>3. phase1.prs
> IP = intonational phrase
> AG = accent group
> foot = foot
> ipa = IPA transcription
> stp = stress pattern
> lex = lexical item
> [ = beginning of phonetic word
> ] = end of phonetic word
>
>Comments welcome.
>
It looks very helpful -- and you've done it so quickly!! Well done! I
haven't checked through every example, but in general the dictionary has
made a pretty good job of the IPA. At this stage it's the stressed
syllable transcription which is most important, and hopefully, when we get
our spoken version, the pronunciation won't be too different from the
dictionary's. I checked one example where I wasn't sure what the
dictionary would recommend -- "renege" -- and unfortunately it prefers /eI/
to my preference, /e/ (short vowel), but we can either hand-correct it or
throw it out.
It is the characteristics of the stressed syllables which are likely to be
most crucial to retrieval. We are going to need therefore to be able to
search for e.g. VV rimes (long vowels and diphthongs), heavy syllables,
plosive onsets -- whatever natural class is important to a particular
analysis -- and we'll need advice from York about how best to express and
represent this within the structure.
I have a query about the representation of the initial unstressed
syllables. I note that they appear as weak branches of a foot/AG which has
no s branch. This is fine (it was what we thought we'd have to do); when
searching for IPs with a specific number of AGs/feet, will it be
straightforward to search for feet containing (s), and eliminate the weak
ones from the count?
Presumably, this phenomenon applies exclusively to IP-initial phenomena?
If any of the items in group 1 find themselves preceded by another "real"
foot/AG within an IP, they will be correctly attached to the earlier
stressed syllable.
I think this ought to work...
Many thanks for the files. I feel encouraged that we have a usable
representation that allows us to cross-reference to lexical and syntactic
information successfully. Now we need to plug in York's bit, if possible!
In due course I'll have to add attributes to AGs, and probably further ones
to feet.
Best wishes,
Jill