Daniel Yacob and Yitna Firdyiwek,
Indiana University, IN., University of Virginia, VA
yacob@apollo.aoe.vt.edu, ybf2u@curry.edschool.virginia.edu
Dedicated to the work and memory of Abraham Demoz (1935-1994)
In the time since the original publication of our paper in "The Journal of Ethio Science & Technology" Volume 3, Number 1 on the topic of representation of Fidel in 7-bit ASCII, the need became apparent to extend the system to encompass representation for Ethiopic numerals, punctuation, and mixed script notations. In the same period more was learned about the treatment of certain characters outside of Amharic that allowed for simplification of the ASCII representation.
The following is a recapitulation of the original publication and an assessment of some of the more recent developments. A complete discussion of many of these changes (and a revised copy of the original publication) are available at the Rensselar Polytechnic Institute ftp archive under the file names ETHIOPIC and ADDENDUM.
As we have indicated before, this system, though well developed, is still not in its final form. Further refinements will only come after many have had the chance to use it and test its strengths and weaknesses on their own. As Abraham Demoz, to whom we have dedicated this work, noted:
"...script reform calls not only for a competent professional assessment of the technical aspects of the script but also for a careful weighing of these against the psychological and socio-political factors that have a bearing on the written word and all that it stands for" (Demoz, "Amharic Script Reform Efforts". ETHIOPIAN STUDIES. S. Segert and A.J.E. Bodrogligeti, Eds. 1983).
Any and all feed back will be greatly appreciated.
PART I
1. The System for Ethiopic Representation of ASCII (SERA) Table
Although some questions still remain to be answered regarding the number of "forms" to use for the ASCII/ETHIOPIC table, we have retained the original arrangement of twelve (12) for SERA pending decisions relating to the Unicode/ISO standards currently under discussion. (See Abass' paper in this conference.) We do not believe a change in the matrix of the table will affect the work discussed in this paper.
The Ethiopic Script in ASCII
1 2 3 4 5 6 7 8 9 10 11 12
g`Iz ka`Ib sals rab`I hams sads sab`I diquala
1 he hu hi ha hE h ho
2 le lu li la lE l lo lWa
3 H2e Hu Hi Ha HE H Ho
4 me mu mi ma mE m mo mWa
5 `se `su `si `sa `sE `s `so
6 re ru ri ra rE r ro rWa
7 se su si sa sE s so sWa
8 xe xu xi xa xE x xo xWa
9 qe qu qi qa qE q qo qWe qWu qWi qWa qWE
10 Qe Qu Qi Qa QE Q Qo QWe QWu QWi QWa QWE
11 be bu bi ba bE b bo bWa (Q is Tigrignia)
12 ve vu vi va vE v vo vWa
13 te tu ti ta tE t to tWa
14 ce cu ci ca cE c co cWa
15 `he `hu `hi `ha `hE `h `ho hWe hWu hWi hWa hWE
16 ne nu ni na nE n no nWa
17 Ne Nu Ni Na NE N No NWa
18 e\a u\U i A E I o\O e3(e3 as in "e3re!")
19 ke ku ki ka kE k ko kWe kWu kWi kWa kWE
20 `ke `ku `ki `ka `kE `k `ko (`k is Chaha)
21 Ke Ku Ki Ka KE K Ko KWe KW KWi KWa KWE
22 Xe Xu Xi Xa XE X Xo (X is Chaha )
23 we wu wi wa wE w wo
24 `e `u `2 `a `E `I `o
25 ze zu zi za zE z zo zWa
26 Ze Zu Zi Za ZE Z Zo ZWa
27 ye yu yi ya yE y yo
28 de du di da dE d do dWa
29 De Du Di Da DE D Do (D is Oromiffa)
30 je ju ji ja jE j jo
31 ge gu gi ga gE g go gWe gWu gWi gWa gWE
32 Ge Gu Gi Ga GE G Go (G is Chaha)
33 Te Tu Ti Ta TE T To TWa
34 Ce Cu Ci Ca CE C Co CWa
35 Pe Pu Pi Pa PE P Po
36 Se Su Si Sa SE S So SWa
37 `Se `Su `Si `Sa `SE `S `So
38 fe fu fi fa fE f fo fWa
39 pe pu pi pa pE p po
(Letters will be referred to both by their ASCII spelling and by their position on the above number matrix (e.g. "he" or 1/1). The columns arealso known as "forms" (e.g., first form, second form, .etc.) or by their Ethiopic names: e.g. g`Iz, ka`Ib, sals, . . .etc.)
2. Considerations We Took in the Development of SERA
We have taken the following two considerations in coming up with our proposed standard
a) The system must be easy to type on a 101 keyboard. This entails:
-- finding the closest match between the Latin and Ethiopic phonetic system (while being as systematic as possible with the inevitable exceptions),
-- limiting the number of keystrokes necessary for each Ethiopic character to a minimum, and
-- placing the most frequently used keys as close as possible to the "home keys" row of the 101 keyboard
b) The system must also be easy for machine translation. In this case, the systematicity of the mapping of Ethiopic to ASCII is exploited to make the machine translation between ASCII and Ethiopic script (in word processors, for example) as fast as possible.
3. Development of the System
It may first occur to one when attempting to write Ethiopic script with Latin letters, to represent the 7 forms with numbers as so:
Consonants:
h1 h2 h3 h4 h5 h6 h7
Independent Vowels:
a1 a2 a3 a4 a5 a6 a7
It is soon found in practice, however, that while this is a very simple system for representing the Ethiopic characters, it is not so simple to read or write in it (e.g., "T5n1y6s6T6l6N6", "a1d5s6 a1b1b4"). This is true largely because our minds are not trained to associate the Latin script with Arabic numbers to form words. One will soon wonder why not use the Latin vowel letters to denote the 7 forms of the Ethiopic characters. This is where the trouble begins: How do you represent the standard 7 Ethiopic forms (plus the "W" forms) with only 5 Latin vowels?
The first step we took was to assign a punctuation mark (the apostrophe ') and "I" for the two extra Ethiopic vowels (plus "W" for forms 8-12). So, following phonetic guide lines we came up with the following system:
Consonants:
h' hu hi ha he hI ho
Independent Vowels:
a' au ai aa ae aI ao
Again, after some trial use (e.g., "Ten'yIsITIlINI", "a'disI a'b'ba") we found that the writing can be made more readable if we used only one character for the pure vowel form. Then the system reduces to:
Consonants:
l' lu li la le lI lo
Independent Vowels:
' u i a e I o
and our sample text would look like: "TenayIsITIlINI", "'disI 'b'ba" which becomes a little easier to read and to type.
After a short time a reader is likely to find that trying to "read a sound" from punctuation proves too difficult. Our minds have been conditioned for too long already to skip over apostrophes when reading possesive and contracted words. We introduce the principle now that whenever possible punctuation be avoided to represent spoken sounds and seek another alphabetic character to replace the apostrophe.
We find a suitable substitute in "E" but recognize right away the draw back of the extra "shift" required to type it. With only a small intuitive feeling one will come to realize that the 5th form letters are used less often in writing than are 1st form. Hence a swap between the two forms makes the use of "E" a little easier and gives us the new table:
Consonants:
le lu li la lE lI lo
Independent Vowels:
e u i a E I o
and our sample text appears a little more naturally as: "TEnayIsITIlINI", "edisI ebeba"
It is at this point that we began to notice two problems:
1) the 6th (or "sadis") form of the Ethiopic characters occurs more often than any other form (about a third more often),and
2) the use of "e" for the first vowel makes the "look" of some familiar Amharic words peculiar, and the sound association is poor.
The quick solution:
1) stop using "I" for the sadis (sixth form) consonants, letting the consonants stand by themselves, and
2) allow the use of "a" for the first form independent vowel with "e", and introduce "A" for the 4th form independent vowel.
Consonants:
le lu li la lE l lo
Independent Vowels:
e\a u i A E I o
Examples:
TEna ysTlN
adis abeba
Indemn kermachWal
zarE Tewat suq hEjE neber
manew smh? manew smx?
Ambiguity Problem with The Independent Vowel
This system is easier to read and type, but there is still a problem. If you have never before seen the word "Tena" how will you know if you are reading 2 Ethiopic characters or 4 -- "TE-na" or "T-E-n-a"? This problem of ambiguity usually occurs because it is not clear whether a consonant letter is a sadis (6th) form followed by an independent vowel form, or a syllable made up of the consonant and following vowel form. Of course, this is a problem only if the reader does not know the language. An Amharic speaker would not make such a mistake.
In another scenario, the name "Gabriel" can be read "ge-b-r-E-l" (correctly), or "ge-b-rE-l" (not quite correct, but okay when speaking fast). Though the ambiguity is there, whether you interpret the Latin as showing 5 (ge-b-r-E-l) characters or 4 (ge-b-rE-l) makes almost no difference.
These conditions may not always be true, however, and the difference does become a big problem for word processors and computer software for translation. It is better then to insure that the characters are unmistakably represented. To accomplish this, our decision was to recycle the apostrophe ' as a separator for independent vowels that appear after a sadis (sixth form) consonant. Thus, we can rewrite Gabriel as "gabrEl" and modify our system, which now includes a third category, accordingly:
Consonants:
le lu li la lE l lo
Independent Vowels:
e\a u i A E I o
Independent Vowels Following a 6th Form Consonant:
l'a l'u l'i l'A l'E l'I l'o
l'e lU lO < --also
If we consider now an application for the remaining uppercase vowels; "U" and "O", we find that in some instances, as shown in the 2nd row of the third category, the use of the apostrophe may be omitted without confusion.
4. Some Commonly Asked Questions
1. Why not use "sh" for "x" and "ie" or "y" for "E"?
These would make logical choices for readers familiar with rules in English but may not make sence in non-English speaking nations where a form of the Latin script is used. It is desirable also to keep the keystrokes to a minimum for humans, the parsing requirements of computers as simple as possible, and media and transfer sizes to a minimum by avoiding multiple character representations when possible.
Further, the reader is left to infer the meaning "sh" as one or two Fidel characters. The separator ' presents a solution here but again complicates parsing and introduces special case rules vs generalized ones. The acception to the general rules also lends towards greater occurences of spelling errors. "ie" may be an easier keystroke than "E" but again introduces inference and parsing complexity. The choice is not always logical as a phonetic model for the "ay" sound with Latin letters when considering such examples as "die", "vie", "pie", "lie", and "tie".
"y" occurs more commonly in speech and written text as a consonant than as a syllabic form. Hence the lowercase Latin character is better reserved for the consonant to save on keystrokes.
When an Ethiopic interface is available, these kinds of questions become input method issues and not file IO and transfer which SERA was primarily designed for.
2. What if I wish to show more sound for a sadis consonant?
* It is not always accurate to say that the vowel component of the sadis consonant is not spoken. For many words the vowel in the 6th form consonant is clearly enounced. If you wish to write in a more phonetic manner with out loss of clarity; this may be accomplished by writing the 2 character representation form of the sadis consonant when it is needed. As you will recall we have redefined the 2 character form of the 6th consonant as "l|". We can mix the two character and one character forms together in the same word to show when the vowel portion is voiced:
ysTlN = y|sT|l|N
tgrNa = tg|r|Na
alfelgm = alfel|g|m
TrE = T|rE
Writing with both the one and two character representations of the 6th form consonant together may be more laborious to the typist but has the advantage of giving the reader a better demonstration of the word's sound when spoken. The mixed representation is not ambiguous and does not pose any problem for machine translation when going from Latin to Ethiopic. If it would become a common practice to mix the two systems, we may wish to try alternate characters in place of the pipe ( "|" ).
3. I see the ' used in other ways, what are the complete rules?
The apostrophe was introduced as a separator to indicated that a vowel after a 6th form consonant does not modify the form of the consonant, ie "nE" is one Fidel letter and "n'E" is two. The principle of the separator may be applied elsewhere when it enhances clarity. For instance between vowels as in "beadis" vs "be'adis" or "keityoPya" vs "ke'ityoPya". Here, the ' helps prevent the reader from slipping back into rules of English where the vowels would be combined into a single sound. Also ' following a consonant as in "t'" may be treated as another definition for the 6th form representation when convenient.
4. Why Are Numbers Used With Letters?
A problem that occurs when trying to represent Ethiopic script phonetically in Latin is the presence of Ethiopic letters that are phonetic equivalents. These cases are encountered with the two Ethiopic characters for "s" and "S" and the 4 characters for "h". Representing one of the 2nd forms with an unused Latin character, say F, R, or V, would be a digression from phonetic norms and adds a level of complication to the reading. In the case of what would be h4 the uppercase "H" is chosen for representation. This choice models the husky "kh" sound that the character has in Tigrinia and other languages.
For the more common type of email exchanges omitting the number 2 or 3 does not result in a loss of interpretation. The use of the ordinals becomes more important later if the text is to be read and translated into Ethiopic script by computer.
5. Why Does "s2" Come Before "s" ?
The "2" is only needed to distinguish the difference between the two "s"s in Ethiopic script. In modern writing it is the newer "s" (the 2nd "s" appearing in the fidel) that finds the most frequent use in the spelling of words. The first "s" is represented as "s2" because it occurs less frequently in writing. Were the 2nd "s" labeled as "s2" it would give the typist considerably more finger work to perform.
6. How was "e3" arrived at for the 8th vowel?
The choice of "e3" is thought to be the best model for the sound of the character. The choice of a numeral to follow "e" will detract from the reading quality of the character, which should come at a small cost when its infrequent use is considered.
7. Why is The Capital "W" Used For Diqala Forms?
The uppercase "W" is used to remain phonetically consistent with the sound of the diqala forms (forms 8 - 12). The lower case "w" is reserved exclusively for consonant 21 with the "w" sound.
Thus confusion and ambiguity is avoided with use of the uppercase "W".
8. Why is "Wu" Used For the Letter I learned was "W"?
Actually both are valid under SERA. In different geographic regions, and at different times within the same region, people have been taught two different sounds for the 2nd form labiovelar (which one may have learned as a 6th form). Phonetic representation as "kWu" "kW" and "kW'", in example, is permitted for both ways a person may have been taught.
9. Why is "hWa" used in place of "`hWa" or "h2Wa"?
This is a break in consistency from how forms 1 through 7 of "h2" were represented. However, as "h" does not have forms after the sabI (the 7th form) there is no opportunity for confusion to arise from the omitted "2" of "h2W". Hence "hW" will be uniquely identifiable as representing diqala forms of the h2 consonant. The advantage of dropping the "2" in the diqala range, will be the keystroke saved for typists.
5. A Full Sample Text with Statistics
WORD COUNT : 170
CONSONANT COUNT
Form1 : 161 Form2 : 21 Form3 : 35 Form4 : 106
Form5 : 14 Form6 : 216 Form7 : 25 Form8 : 3
VOWEL COUNT
Form1 : 25 Form2 : 0 Form3 : 5 Form4 : 2
Form5 : 1 Form6 : 13 Form7 : 1
From the Ethiopian Examiner January 1994
yeselamna yeIrqu konferans gizEyawiw mengst keslTan Indiwerd Teyeke
bekefateNa gugut siTebeq yeneberew yeselamna ye`Irq gubae, ketahsas 9-13 1986 `a.m. beadis abeba ketema baderegew yeamst qen sbseba, beih`adEg yemimeraw gizEyawi mengst slTanun Indiyasrekb Teyeqe.
qedem blo paris lay sbsebaw Indidereg keTeyequt sebat teqawami budnoc wsT, yesostun abalat wede ageracew sigebu awroplan Tabiya lay bepolis bemasyazna bmaser mengst bzihu sbseba lay Indaysatefu adrgWacewal. yetasrut abalat, ato abera yemaneab, we/rit genet grma, ato mesfn tefera, ato alemayehu dErEsana ato genenew asefa (keidE`haq): ato seyum zenebe (kemed`hn) Ina ato ibsa gutema (keoneg) nacew.
mengst Inezihu sewoc lay yewesedew yeIsrat Irmja sewocn beselamawi menged beageracew yepoletika hidet wesT Indaysatefa slemiyaderg bzu sewocn aseqoTtWal. beadis abeba yemigeNu diplomatocm yKEw yemengst Irmja yesbsebawn tesatafiwoc farhat lay bmeTal sbsebaw mnm bego wTEt IndayameTa yaderg yhonal bemalet hesabacewn gel`Sewal.
yeityoPya gizEyawi mengst (iH'adEg) besbsebaw lay saysatef qertWal. lezihum begizEyawi prEzidEntu beato meles zEnawi yeteseTew mknyat sbsebaw lepropaganda `alama bca yemidereg kentu sbseba new yemil new.
The following is an annotated list of appendixes that address various other aspects of SERA. The complete texts are
6. Other Appendixes Not Included in this Version (as noted at the begining of this discussion complete and up to date copies of all of these texts can be found at the Rensselar Polytechnic Institute ftp archive under the file names ETHIOPIC and ADDENDUM.)
a. Questionnaire and Sample Text
b. On The Character Specific Representation of Numbers:
c. Generalized Pseudo-Code for Latin --> Ethiopic Translation
d. Ethiopic in Emacs