The following is a cut-and-paste copy of the definition of the IPA/ASCII standard, widely used on sci.lang (among other places) for accurate phonetic transcription in a 7-bit medium. Aside from splitting off the large appendices into their own nodes and hardlinking some of the more esoteric linguistic terminology, I have left it untouched, since its author knows the subject far better than I do. The page was written by Evan Kirschenbaum and can be found at http://www.hpl.hp.com/personal/Evan_Kirshenbaum/. I'm not expecting upvotes for this, but please don't downvote it either -- I refer to IPA/ASCII in a number of nodes on the Japanese language, and I expect that quite a few other linguistics-oriented noders will find it helpful. --gn0sis


Representing IPA Phonetics in ASCII

Evan Kirshenbaum <kirshenbaum@hpl.hp.com>

Last Modified, 4 Jan 1993 / Error corrected 22 Jan 2001

This article describes a standard scheme for representing IPA transcriptions in ASCII for use in Usenet articles and email. The following guidelines were kept in mind:

  • It should be usable for both phonemic and narrow phonetic transcription.
  • It should be possible to represent all symbols and diacritics in the IPA.
  • The previous guideline notwithstanding, it is expected that (as in the past) most use will be in transcribing English, so where tradeoffs are necessary, decisions should be made in favor of ease of representation of phonemes which are common in English.
  • The representation should be readable.
  • It should be possible to mechanically translate from the representation to a character set which includes IPA. The reverse would also be nice.
In order to be able to represent a wide range of segments while making common segments easy to type, we allow more than one representation for a given segment. Each segment has an "explicit" representation, which is a set of features between curly braces ("{" and "}"). Each feature is represented as a three letter abbreviation taken from a standardized set. The phoneme /b/ (a voiced, bilabial stop) could be represented as /{vcd,blb,stp}/. A first cut at the feature set appears in appendix A.

The word tag could thus be represented phonemically as

/{vls,alv,stp}{low,fnt,unr,vwl}{vcd,vel,stp}/
and phonetically as
[{vls,asp,alv,stp}{low,fnt,lng,unr,vwl}{unx,vcd,vel,stp}]

This works, but it's a bit of a pain. To simplify transcription, we allow an "implicit" representation for a segment which consists of a (generally alphabetic) symbol followed by diacritics. Thus /b/ stands for /{vcd,blb,stp}/. Case is significant (/n/ and /N/ are different segments). The segment symbols are given in appendix B.

The word tag can thus be represented phonemically as

/t&g/

The diacritics for a segment are represented between angle brackets ("<" and ">") and consist of symbols or features. (In the common case where the diacritic symbol is a single character which does not encode a segment, the brackets may be removed.) The features which the diacritics map to override those of the segment.

The word tag thus becomes narrowly

[t<asp>&<lng>g<unx>]
or
[t<h>&<:>g<o>]
or
[t<h>&:g<o>]

Some diacritic symbols encode more than one feature set. Which one is meant should be apparent from context. For example, "." stands for "{rnd}" when attached to a vowel, but "{rfx}" when attached to a consonant.

Clicks are common to many languages (especially in Africa), but there is no IPA diacritic that means "click". Rather than use up several characters for clicks (which are infrequent in the languages most often discussed), we instead use the diacritic "!" after the homorganic unvoiced stop. Thus /t!/ (= /t<clk>/ = /{alv,clk}/) is the sound commonly written tsk and used in English to show disapproval.

The complete set of diacritic symbols appears in appendix C below. Appendices D and E contain representations of segments more or less ordered by feature (appendix D in tabular form, appendix E as a list). Appendix F contains a list of all of the ASCII characters and the uses they have been pressed to.

For transcription of any specific language a group can by convention alter the character mappings (as an example, for Spanish /R/ may be better used to represent /{alv,trl}/ than /{mid,cnt,rzd,vwl}/). An author may also press a little used symbol (for the language under consideration) into service to highlight a distinction. Such an alteration should be made explicitly to avoid confusion.

The diacritics "+" and "=" and the segment symbols "$" and "%" are explicitly left unspecified so that they can be used to mark language-specific features (that are otherwise cumbersome to mark). Such symbols can be assigned either by convention for a specific language or in an ad-hoc manner by an individual author.

Stress marks are prepended to the syllable they attach to. "'" signals primary stress, "," signals secondary stress. Spaces should be employed to separate words (cliticized words may be written unseparated). When discussing single words, it may be helpful to insert a space before each syllable that doesn't carry a suprasegmental marker.

Thus, I hear the secretary for an American might be something like

/aI hir D@ 'sEkrI,t&ri/
while to an Englishman it might be more like
/aI hi@ DI 'sEkrVtri/

Transcribing tone is harder. Here's an attempt. For register tone languages (e.g., Hausa, Navajo), numbers should be used with one being the lowest. Thus in Navajo, "1" is low tone and "2" is high. In Yoruba "1" is low, "2" is mid, and "3" is high. The language's "default" tone need not be specified. For contour tone languages (e.g., Mandarin, Thai), there is generally a numeric system in place (Mandarin: "1" is high, "2" is rising, "3" is falling rising, "4" is falling). The tone indication should follow the syllable (vowel?).

The symbol "#" is used to represent a syllable or word boundary.

Appendices

Appendix A. Feature Abbreviations
Appendix B. Segment Symbols
Appendix C. Diacritics
Appendix D. Segment Table
Appendix E. Segment List
Appendix F. ASCII Table