perm filename SYMBOL[4,ALS] blob sn#059722 filedate 1973-08-24 generic text, type T, neo UTF8
00100		These are some random thoughts on the problem of a satisfactory
00200	code for acoustic segments.
00250	
00300		We need to recognize several levels in our segment classification.
00400	In the first place we should be able to specify such broad groupings as
00500	VOICED, FRICATION, and STOPS. Then we want to be able to tag subdivisions
00600	such as VOWEL GLIDE and NASAL. Finally we want to be able to specify
00700	specific segments in as much detail as is possible and quite independent
00800	of the usual phonemic classification. That is, we should be able to 
00900	identify transition regions in vowels preceding stop consonants, the
01000	stop, the burst, the aspiration and the following vowel transition, all
01100	as related parts, perhaps independent of the actual preceding or following
01200	vowel. At the same time we should be able to classify a vowel transition
01300	as an entity by itself quite without regard to its association with a
01400	specific consonant or vowel.
01450	
01500		At first glance it seems that we should use symbols that are 
01600	related to the usual phonemic symbols or to orthographic symbols for
01700	the general type of sounds. This does cause some confusion, however, and
01800	it might be argued that we should invent a quite independent set of
01900	symbols that will never be confused with their phonemic counterparts. An
02000	independent set poses a problem of learning to use it and of resistance to
02100	its use by people with linguistic or phonetic backgrounds. Several possible
02200	types of symbols with some intrinsic meaning might be proposed. The
02300	trouble with this suggestion seems to be that the symbols would have to
02400	each consist of several letters for them to be easily recognized and
02500	this poses a problem of their use on charts and in tables. Also if they
02600	are simply abbreviations of names there would be a tendency for users to
02700	invent abbreviations that are not in the accepted list with the attending
02800	confusion.
02900	
03000		Another aspect of the problem has to do with the way data is to be
03100	stored in the header which is associated with the acoustic data itself.