Converting FLEWSY to another system

This page discusses a number of issues that frequently come up in trying to convert the FLEWSY notation to an alternative spelling system.  Everything on this page is just my opinion, and you are of course completely free to ignore any parts of it with which you don't agree.

This page is oriented towards spelling-system developers, and discusses issues of representation of sound, readability, consistency, and the like.  Programming issues are discussed on this page.

When to use FLEWSY

• Your system may depend on information which is not recorded in FLEWSY, such as syllable boundaries or word origins.

• You may disagree with the pronunciations shown in FEWL.  For instance, FEWL shows the first vowel sound in the words behave, precede and refine as a short i sound.  One school of thought holds that it is more consistent to spell these words to indicate a long e rather than a short i.  There is no way for the conversion process to determine which short i's should be handled this way.

• FEWL shows many words having more than one syllable with primary stress.  Some systems make the (fairly reasonable) assumption that only one syllable can have primary stress.

In these cases, too, it may be possible to do a FLEWSY conversion based on simplifying assumptions, and then repair the dictionary manually.

Finally, there is one other requirement that might seem obvious, but really needs to be stated.  Conversion from FLEWSY to another notation requires a program.  Programs aren't vague, and can't be coded by waving your hands and saying "You know what I mean".  To do a FLEWSY conversion, you will have to understand your own system and its rules thoroughly, and there will have to be rules.  If you are a programmer, you will have to understand how you intend your system to work well enough to translate it into code, and if you aren't, you will have to be able to explain it to a programmer who won't have the insights you have that make everything so obvious to you.  There are probably many systems for which FEWL and FLEWSY are of no use whatsoever, simply because these systems are too imprecisely defined to allow a converter to be written.  This is perhaps a shame, but in my opinion work spent on making such systems well-defined and comprehensible is unlikely to be wasted.

Unstressed vowels

Many of the most difficult issues associated with alternative spelling systems have to do with the representation of unstressed vowels.  There are two great dichotomies here, and the denizens of Saundspel can be found arguing about them any time of day or night.  These are accuracy vs. adequacy, and predictability vs. familiarity.  FLEWSY was designed for use by both sides of these great debates.

The proponents of accuracy believe it to be very important that the spelling reflect a word's pronunciation exactly, while the proponents of adequacy would say that only a "good enough" match to the pronunciation is required.  In particular, the adequacy advocates would argue that, so long as you get the stressed syllables of a word right, how you spell the unstressed ones is not so important.  The adequacy guys would say that the accuracy guys want the spelling to split hairs, while the accuracy folks accuse the adequacy folks of producing incorrect and/or unpredictable spellings.

The proponents of predictability believe it is very important to be able to predict how a word will be spelled with hard and fast rules, even if most of the spellings are very unlike TS.  The proponents of familiarity would argue that resemblance to TS is a virtue, and trumps predictability when an imperfect but familiar spelling improves recognition.  The predictability faction argue that stressing familiarity keeps spelling hard, while the familiarity faction argues that insisting on predictability makes reform impractical.

I don't bring these issues up to settle them (fat chance of that!), but to note that FLEWSY is a notation which can be used in support of either philosophy.

Perhaps the most obviously contentious issue in spelling system design is what to do about the schwa and the unstressed short vowels.  There are a number of positions one can take here:

• The schwa and the unstressed vowels should be spelled distinctly for accuracy.  This may be accomplished by adding a symbol for the schwa (perhaps ø) or by reserving a vowel letter (say a) for schwa, and spelling the corresponding short vowel in some other way (for example, à or ae).

• The schwa can be spelled the same as one of the short vowels without introducing incompatibility, but it should always be spelled as the same short vowel.  Usually, the vowel u is selected for this purpose.

• The schwa can be represented with any of the five vowel letters, making it ambiguous whether the schwa or a short vowel is intended.  The letter used is generally taken from the traditional spelling.  This position strongly emphasizes familiarity at the expense of predictability.

The FLEWSY representation for the schwa makes it easy to convert to any of the above techniques.  The schwa is denoted in FLEWSY by one of a, e, i, o, u, µ or 3.  Proponents of one of the first two positions above can convert all these to the schwa symbol, while proponents of the third can retain the FLEWSY spellings except for µ and 3.

There is another angle to the whole schwa question, one that some reformers are unaware of.  This is the problem of short i.  There are a large number of English words containing an unstressed vowel which is not clearly a schwa, and not clearly an unstressed i.  The i in "ratify" is a good example.  Some people say /r{tIfaI/ and some say /r{t@faI/, and it is very hard to hear the difference.  The various dictionaries are themselves unable to agree on which sound is more correct or more commonly used.  I will call this sound the "indistinct i" for the rest of this discussion.

There are at least three ways this issue can be settled:

• You can always spell the indistinct i as a schwa.  This can produce quite unfamiliar spellings, like wikud or kiendnus.

• You can always spell the indistinct i as a short i.  This too can generate unfamiliar spellings, like kandidit or glasiz.

• You can spell the indistinct i as in traditional spelling.  This solves the familiarity problems, but then the usual predictability problems surface.  Why should kwaality be spelled differently from gayety?
  

Again, FLEWSY is very accomodating.  The indistinct i is spelled as one of ä, ë, ï, ö or ü.  Proponents of the first two positions above can translate them all to the appropriate symbol; proponents of the third can just drop the umlauts.  Note that if your normal representation of the schwa is the letter i, the first two positions are pretty much the same.

There is a third extremely common predictability/familiarity conflict in many English words, namely those in which an a or e is pronounced as an unstressed i, such as "vintage" or "entrust".  FLEWSY uses â or ê in such words, again giving one the flexibility of choosing whether to retain the traditional spelling or use a less familiar phonemic one.

Dialectal Issues

Another class of issues for conversion of FLEWSY spellings has to do with the existence of multiple pronunciations for words, and in particular of British/American differences.  I should note that FLEWSY, and everything about it, was designed for American English.  Nevertheless, some people believe that spelling systems can be designed in which most words are spelled in a way acceptable to British speakers as well as American ones.  The primary way in which FLEWSY supports this faction is by distinguishing the broad a sound in "father" from the short o sound of "pop".  These two sounds are the same to American speakers, but different for mainstream British speakers.  Some would argue that a single spelling should be used for this sound in the interests of accuracy and predictability, and the British can take care of themselves.  Others would claim that having two spellings for this sound isn't all that bad, and contributes greatly to having a single good orthography for all forms of English.

Again, it is certainly impossible that I could settle this dispute.  FLEWSY uses both A and ò for this sound, depending on the British pronunciation.  FLEWSY converters need not translate these letters distinctly if the target system uses a single representation.

A similar issue is that of the long u, which is often used to refer to two different sounds: /ju:/, as in "cue", and /u:/ as in "true".  Some words have both pronunciations, such as "due" and "Tuesday".  This is in part a British/American difference, but not entirely, since some American speakers favor the /ju:/ pronunciations.  There are at least three positions a reasonable system might take here:

• Spell both sounds the same, e.g., as ú.

• Spell the two sounds differently, and prefer the /u:/ pronunciation when there is a choice.

• Spell the two sounds differently, and prefer the /ju:/ pronunciation when there is a choice.

FLEWSY spells /u:/ as U, and /ju:/ as yU.  When both pronunciations are commonly used, the notation y?U is used.  Thus, any of the three options above can be supported.

FLEWSY uses similar notations in other situations where speakers vary in whether or not particular sounds are spoken.  For instance, *l is used to represent either l or al (as in fìzìk*l¥), *w is used to represent either w or hw (as in *wìmper), *y is used to represent either y or ¥ (as in  jIn*yus), *X is used to represent either X or C (as in tèn*Xon), and ? can be used after any letter to show that the sound is optional (as in Yvo?r¥).  As usually happens, one can have one of several philosophies here:

• Some prefer to show such optional sounds in the spelling, believing that it is better to sometimes have silent letters than to have sounds left out of the spelling, that is, that jenurul is a better spelling if you don't pronounce the second vowel than jenrul is if you do.  Some also prefer to show these sounds because they believe that doing so encourages precise speech.
  

• Some prefer to leave out the optional sounds to make spellings shorter, or because the shorter pronunciation is generally the majority pronunciation.

Again, FLEWSY easily supports either approach.  A complication is the fact that one might not to treat all words of the same sort the same way.  For instance, one might prefer in general to leave optional sounds out, but not in the words "virulent" and "liberal".  Special cases like this are best handled by editing the dictionary after it has been converted, unless the exceptions can be systematically recognized by programming during the conversion.

The ? character appears most frequently in FEWL following y (in forms of long u), d (in -and- words), and schwa or indistinct i (before l, n or r).  You probably want to handle these uses of ? consistently, either always leaving the preceding character in, or always removing it.  There are a number of other uses of ? which occur in only a few words, such as klOD?z (clothes), h?&bal (herbal), màdemw?azèl (mademoiselle), etc.  I recommend that these question marks be allowed to remain in the converted output, for later examination and manual resolution.

Finally, I ought to mention a specific case of variant pronunciation which is not specially encoded in FLEWSY, but which is easy for a program to recognize.  This is the case of the /{r/ sound which is often used in words like "marry" and "apparent", which are equally often pronounced with /Er/.  This is encoded in FLEWSY as àr.  Those who feel this sound should not be distinguished from /Er/ can treat this sequence the same as èr.

Morphological issues

The English lexicon is not just a collection of simple one and two syllable words.  It is complicated by inflections, prefixes, suffixes and compound words.  In my opinion, many systems which do fine with simple root words fall apart when faced with these morphological difficulties.  The issues are not easy, and using FEWL and FLEWSY will not force you to face them (much less to come up with a good solution), but it at least affords you the opportunity to try, and to discover whether your solutions actually work.

As with the issues above, there are two schools of thought here, which we may term structuralist and phonemicist.  The structuralists believe it is more important to be structurally consistent than it is to be phonemically consistent; the phonemicists disagree.  I should make my own alignment clear here - I am a structuralist.  This seems to be a minority position among reformers.

The most straightforward morphological issues concern how to handle the past tense and plural inflections, indicated in TS by -(e)d and -(e)s.  For each of these inflections, there are two positions one might take:

• One can hold that the pronunciation should be indicated correctly - plurals should be indicated by -s or -z, as appropriate, and the past tense should be indicated by -d or -t, as appropriate.

• Alternately, one can hold that the plural and the past tense should always be indicated by the same character, regardless of pronunciation - s for the plural (or z if you're a wyrdo like me) and d for the past tense.

Something I find rather surprising is that there are systems which treat these two cases differently, as by always using s for the plural, but both d and t for the past tense.

In any case, FLEWSY supports all these choices.  FLEWSY indicates a plural ending with $, and a past tense with þ.  Since the actual sound of the inflection is very easy for a program to determine, it is simple to translate these special symbols appropriately, according to your preference.

The prevalence of English affixes presents many similar issues.  In a nutshell, many prefixes and suffixes vary in pronunciation from word to word and speaker to speaker, often in a far less predictable manner than -ed.  For instance, the prefix anti- is sometimes pronounced /{ntI/, sometimes /{nti/ and sometimes /{ntaI/.  The prefix bio- is sometimes /baIoU/ and sometimes /baI@/.  The suffix -man is sometimes /m{n/ and sometimes /m@n/.  Structuralists believe that these affixes should be spelled the same, regardless of their pronunciation; phonemicists disagree.  FLEWSY is only of limited help here.  It uses the / character to mark many common prefixes, and ( to mark those suffixes whose pronunciation is variable.  (The support is limited because it have been a lot more work to have done any better, and because many suffixes change the pronunciation of the root word in ways that are difficult to deal with systematically.)  If you are a structuralist like myself, the best I can say is that FLEWSY will give you a start on tackling the problem, and the rest is left to manual editing and your patience.  And if you're not a structuralist, FLEWSY will not get in the way of your representing affixes exactly as they are pronounced.

Finally, something should be said about compound words.  Compound words are an inherent and important part of the English language.  In TS, compound words are usually formed by just adjoining the components (as in carpool and somewhere), though sometimes, more or less at random, a hyphen is used to separate the parts (cold-blooded, know-how).  The problem is that compounding sometimes causes changes in pronunciation or stress which would be reflected in a reformed spelling, making it harder to recognize the parts of the word.  For instance, the normal pronunciation of "grandmother" leaves out the d, but leaving it in the spelling makes the components of the word clearer.  For another example, in my system WLM, the schwa is represented by an a at the end of a word, but by an i within a word.  The English word "sofabed" would be spelled phonemically in WLM as sofibed, which doesn't look particularly related to sofa.

The easiest solution to such problems is probably to just write all compound words with a hyphen.  But this is not a solution many are likely to favor.  It is more natural in American English to leave the hyphen out than to put it in, and many words that used to be hyphenated are now most often spelled unhyphenated, as with "babysitter".  Another solution is to write all these words phonemically, but even hard-core phonemicists are unlikely to accept shountel for "show-and-tell".

FLEWSY uses the symbols -, =, /, \ and } to indicate various sorts of word combinations.  (See the FLEWSY definition for details.)  These symbols allow a program to recognize where words are joined, and therefore to change the spelling when necessary.  It must be noted that these markings are very subjective (is "grandmother" really a compound word?), and must be regarded as approximate at best.  If you care deeply about how compounding is reflected in your spelling system, some manual editing of the results of conversion will almost certainly be required.

To comment on this page, e-mail Alan at wyrdplay.org