Lawd! . . . I got them ascii blues.

Transcription troubles

One immediate difficulty for a computational phonologist - unlike a syntactician - is his or her need for a stock of symbols to represent sound segments (phonemes and phones) which the standard computer keyboard and the standard character codes do not directly provide. Not only that, the particular implementation of the programming language which is being used may impose even further constraints. Fortunately, the problem may be worked around in a variety of ways by a Logo programmer.

Co-opting characters

One obvious fix, when you can't find the symbol you would like to use, is to co-opt the nearest look-alike in the available character set. In Logo, for example, 3 is a character as well as a number, and if you feel that b3d reminds you well enough of the transcription you usually adopt for bird, then just borrow 3. (We did just this in setting up the (limited) sonority scale for the simple syllabifier.)

While you will be pushed to implement a universal phonetic alphabet, it will usually be possible, especially when an extended character set is available (using the option or alt key and sometimes key sequences), to find enough mnemonically appropriate characters to satisfy the requirements at least of one language at a time. (Mac users will find that option b provides a good approximation to the IPA symbol for a voiceless palato-alveolar fricative in most fonts.)

But be warned. You need to watch out for potential clashes with characters which have a special status in the programming language. Although [ and ] may have enough similarity to the IPA symbols for a pair of mid-open front and mid-open back vowels to make them candidates for transcribing say pot as p]t and bed as b[d, you would immediately fall foul of their use in Logo to mark the beginning and ends of lists.

Using di-graphs

The co-option tactic is based on the assumption that procedures for processing phonological structures are more easily and efficiently defined when each element of a structure is represented by a single character. Co-opting existing look-alike characters is a way of guaranteeing this relationship at the same time as achieving a readable representation.

It does not follow, however, that, at the point of user input, sequences of characters cannot be used. As long as some pre-processing of an input string converts character sequences where necessary to single character equivalents before the string is passed on, the basic procedures can continue to be defined simply. And it does not matter to them whether the single characters arrived at in the re-coding have mnemonic value or not. If the pre-processor translates the digraph ch (which you might not unreasonably use for the first consonant of chips) into the number 4, the Logo interpreter is not going to bat an eyelid.

Naturally you will need to select with great care the character sequences which you will use to represent single segments. If you think that th is a nice way to handle the dental fricative at the beginning of think (because it matches the orthographic convention) then what about fathead, fathead?

On top of these problems you then have to implement the pre-processor and you will no doubt find that its manipulations slow down the overall processing time. But the trade-off may be worth while. Look here for a simple example of a digraph pre-processor.

Oh, lucky ones!

You just might be lucky enough, of course, to be using an implementation of Logo - like Logo PLUS - which allows you to select the font in the system windows. This is, however, not always the perfect solution it might seem. Some fonts developed for use by phoneticians use character codes which conflict with the codes used for the normal display characters or worse with those used by special Logo characters (like the double quote used to identify words or the square brackets used to delimit lists). In these circumstances, a much needed character may be simply unusable and a printout of a procedure definition might as well be in double Dutch.

The problem of undecipherable Logo code disappears, of course, if separate user input and display windows can be opened, but clashes with special symbols can only be dealt with by careful selection of a font. Of the freeware Macintosh fonts I most often use, PalPhon seems to work more easily with LogoPLUS - for my purposes - than SIL IPA or TechPhonetic. Click here for a page of links to sites holding phonetic fonts.

Ron Brasington
Department of Linguistic Science
The University of Reading