I'm creating a combined bangla/hindi/sanskrit diacritic font - and working out what characters I need to include

Hello everybody

During the recent visit of my (sort of siksha-) guru Sri Raman Bihari Das Babaji, we discussed compiling a song book of all the songs that are sung in our group and also to document my paramgurudev's padavali kirtan that has remained untranslated in devotees' tape collections.

My style of transliteration is somewhat different from the usual type as I also include the original bengali script (and in future the devanagari) and my english transliteration aims only to reproduce the pronounciation rather than the spelling which is available in the original script anyway.

This is an example of one of my dodgy translations in progress.

this chart explains the correspondance between the letters, my tranlsliteration and the conventional system (that in my opinion is inadequate for bengali).

I am pretty happy with the bengali system I have developed and although it does have a few disadvantages I can live with them, they are:

1:
the initial vowel is sometimes silent (in spoken bengali) and sometimes sung (in metre when sung) and as yet I have no symbol for an optional final inherent vowel

2:
currently I have only two possible symbols for the inherent vowel either an "å" or an "o" and there are more than two actual sounds in bangla because it is a continuum of vowel sounds, that I can live with though.

The songbook will include songs in bangla, hindi and sanskrit. For sanskrit I will use the standard transliteration system possibly substituting the occasional gy for jn.

I will probably be using Itranslator99 for the hindi devanagari whuch in the main is pretty good but I am not completely happy with the transliteration it produces mainly for the anushvara and chandrabindu so I am looking for an appropriate replacement. Would a chandrabindu work for that?

I am currently working out symbols I need to include in a font that will:

1:
reproduce pronounciation in Bengali

2:
serve as a normal sanskrit diacritic font

and most crucially at the moment
3:
faithfully reproduce pronounciation in Hindi


I am not really very good at hindi, which I sing like a bengali. I would like to ask the help of people with good hindi pronounciation for help in deciding what would be an appropriate choice of characters. I'll write more later.

As an aside I've decided to create a sans serif font as it will have a distinctive look and not look like balaram!

Any comments or feedback welcome.

Ananga Manjari Das
Glastonbury, Somerset, UK

Here is a hindi example with diacritics generated by Itranslator99

thanks Madhava

Actually there's loads of mistakes. If anyone sees any more message me privately as I'd like to limit the public conversation to topics of legibility, typography, pronounciation etc and the book will go through lengthy proofong and editing but thats a seperate issue really.

I loved your photos of your delhi book printing adventure (going off-topic myself!)

I'll have to look at this more closely when I get the time.

The good thing about Balaram (at least a good direction it's headed...) is that it is convertible from the transliterated font to Devanagari. It was never done with Bengali in mind, and as far as I know, no such conversion program has been written. And there are difficulties in the current Balaram system even if we wanted to do that.

Nevertheless: If this hope exists, then the idea of one-to-one correspondence of glyphs is vital. More than one letter having the same correspondence in English transliteration (such as "j" or "sh") causes problems.

So let's forget that for the time being.

(1) I don't care much for the a-dot solution for the short "o" sound. I would prefer using "o" and "ô" or "ö" for the long o sound. "koro" just seems more natural than "kårå". You could then either abandon the a-ä division, using only "a" as you have been doing, or possibly find a real need for it.

Another possible solution I have considered is to use a character from linguistics, which I am sorry I cannot reproduce here for the moment. (Backwards "c")

(2) The a-i, a-u problem would not be entirely solved by (1) above. The current solution of umlaut is the one that I prefer. haïyA rather than haiyA. (In the current Balaram system, this cannot be done, and so BBT books usually have ha-iyA. ha-uk. This is a very bad solution.

Your system takes care of this, as there is no confusion between åi and oi, whereas my proposed solution above would not solve it.

(3) The "w" in häowä seems superfluous. I have been using jaoa haoa, which in English automatically incorporates the "w" sound. The same for when the letter replaces "y" (though I tend to use it) jAiA could be as good as jAiyA

(4) One of the biggest defects of Balaram is its inability to accomodate properly anuswar/candra-bindu. Most current systems prefer using the tilde over the accompanying vowel, like in Portuguese. I don't care much from m-dot for anusvara, but it's not really a big problem. It's less sloppy than "ng." The phonetic character that looks like an n with a hook on it seems the most succinct and natural solution, but regrettably a little too radical.

Those are preliminary comments. I'll have to look more closely later.

The Hindi looks very nice.

Thanks Jagat for your response.

Whatever system I choose to use it'll be a compromise. I'll stick to the a with a circle (which I'm using instead of the backwards c) cos visually it looks like what it is; something between an "a" and an "o". It is clumsy though having to make a distincion sometimes but I think it is better than the "sri guru chaaaaraaaanaaaaa paaaadmaaaa" scenario. Since I'm including the original devanagari/bangla script that should satisfy all the linguo-geeks who should be able to read the original script anyway
.

I'm not so worried about the bengali although I found out that for the word cha(chandrabindu)DAbe that I need to create the somewhat inelegant glyph with made up from a letter "a" a circle/letter "o" above it and a chandrabindu above that. I prefer a chandrabindu instead of a tilda cos people recognise chandrabindus from the letter OM and people are generally better at OMming than they are at deciphering strange symbols.

I next have to decide which symbols I actually need for hindi.

So far the font I have cobbled together from balaram just uses the uppercase letters for the accented letters since there are no actual uppercase letters in the original scripts I thought they were disposable so I shouldn't have to go beyond the 256 and probably not have to go beyond the "lower 128" either if I continue to use uppercase letters for accented letters. These technical difficulties will probably fall into place when I have researched such matters further.

Prabhus, you may wish to look at transliteration tables for both Bengali and Hindi available at



From the page, you can download "Bengali-Assamese-Manipuri.pdf" and "Hindi-Marathi-Nepali.pdf".

Even if we have disagreements with the system they advocate (apparently an ISO 15919 standard for both Bengali and Hindi -- there is a lot of overlap between the transliteration schemes), it may serve as a good basis for discussion. Explicitly state which characters we ought to change, and provide a rationale. For example, the argument could be made that people trained to pronounce the Sanskrit letters through a unified or canonical transliteration scheme are quite likely to transfer their pronunciation habits to Bengali texts. Therefore, we could replace only the characters which have a distinctly different pronunciation in standard Bengali from the pronunciation of the etymologically equivalent character in Sanskrit.

This could be done by creating a font with all the agreed-upon Bengali variants in distinct positions from their Sanskrit etymological equivalents. If Ananga Manjari prabhu then were to encode a Bengali text using this font, the /a/ would be encoded as an /o/ or /[IPA open-o]/ (or possibly as /a-ring/) glyph, depending upon the correct pronunciation context. Similarly for /oi/ ligatures, etc.

Now, if Jagat prabhu receives a text from Ananga prabhu encoded in this way, he could either run a little Perl or Python script to convert all the characters in the text-stream to standard Sanskrit (assuming he wished to do this) or, even better, make use of a second font which has the identical Roman-transliteration character glyph used in Sanskrit, but located in two positions in his font layout grid. The second alternative has the advantage of preserving bidirectional translation among the two character schemes. Those who wished to download texts by the Jagat method would utilize the special dual-location J-font. Those who wished to more closely indicate the pronunciation differences would utilize the richer-glyph A-font.

Looking over the glyph correspondence table approved by the International Standards Organisation, I note that even the Sanskritists may wish to change their transliteration scheme. It seems, for example, that the ISKCON "r-dot-under" for the vowel in Krsna should really be "r-ring-under". This preserves the dotted r for the flapped ra and rha consonants. Other characters with a macron under or two dots under are used for letters which are dotted in Hindi script.

I hope some of this may be helpful in crystallizing an agreement on direction.

Best regards to all,
Ananda das

Much food for thought, Anandaji. I am afraid that I have gotten too familiar with doing things the way they have been and have not been putting much thought into this.

Certainly the ISO system should be given serious consideration, though I recall finding certain defects. I was not all that crazy about the "r"s. But Balaram is seriously deficient, there's no doubt about it.

I'd like to see how Ananga responds. With your expertise and contribution, maybe we could actually get something done.

I have about five or six Bengali fonts that follow different ASCII schemes. There is an ISCII or something that supposedly standardizes the Indian fonts, but Bangla Desh follows a different scheme. The Assamese seem to follow yet another. The "Amar Bangla" fonts that we use on "Banglaword" follow the Bangladeshi scheme, but there are several deficiencies: Quote marks and other punctuation, for one. But also it misses several characters necessary for Sanskrit text in Bengali font.

Just out of total curiosity, and it has nothing to do with transliteration schemes and the like, but will you be designing the glyphs themselves from scratch? I know Balaram and most Indic transliteration fonts are absolutely terrible looking... I consider Balaram to be absolutely unreadable on a screen... at least on any Linux box I've put it on...

I don't have any experience with glyph design, and I know its a bit of an esoteric science, but I think an excellent screen transliteration font is much needed in the community...

Just my $.02

The deficencies in Balaram are pretty much the same as the deficiencies in the Harvard Kyoto so here is my response to the deficiencies of the latter:

http://www.gaudiyadiscussions.com/index.ph...pic=2207&st=15#

Feedback from bengali linguists please HERE not there.

radhe radhe

Ananga

this should make it clearer

and

Other issues

• the avagraha / lupta a-kAr /apostrophe situation needs clarifying
• For maximum portability between platforms, compatibility with simple text editors it would be best to restrict our font to the first 128 available characters in the font.
• Because indian languages do not have capital letters they are not really needed in a transliteration font so we can use the capital letter positions for our accented characters
• Some kind of conversion program needs to be written to turn upper case letters in documents using balaram into lower case so that they can be compatible with the new font
• Ananga likes to use ae ligature (å) and ao glyph (å) and these need to be nasalised
• Ananga suspects that hindi pronounciation requires ai-, au- and possibly an ng ligature, some of which may need nasalising
• I think that's about it
• radhe radhe radhe jay jay jay sri radhe!

Did we ever get anywhere with extending H-K to accommodate Bengali? Capital Y is very sensible and seems to be free in the grand H-K scheme. However, what about .D, and the a:i, e:i, a:u etc.?

I adapted someone's JavaScript code to make this still somewhat buggy converter (look at message # 9 of this thread) and I decided on Y for the onthoshto jo,





a_i, o_u for two vowels in a row because that was what the original coder whose code I adapted had used and it seems sensible especially as it is a basic ASCII character.





I like the idea of P for gauPiya discussions for three somewhat silly reasons: It sounds like "more gopis" which I like, it is almost a letter "R" and it almost looks like a "D" backwards and its a capital greek letter Rho.





I definitely don't like the .D because when I write by hand I write the dot afterwards and anyway the "." is already used for explaining how words are joined together in sanksrit.





I seem to have problems with khondoto at the end of words not displaying proplerly, with numerals and I still need to work out how to get nitya নিত্য by typing in nitya not nitYa.





Another minor issue is deciding whether " ' " is turned into an avagraha ঽ (for sankrit) or an apostrophe (for bengali)