[Project-ideas] GSoC idea for improving the aesthetic quality of Indian Scripts

[Sankarshan Mukhopadhyay]

On Mon, Apr 22, 2013 at 1:06 AM, Mayank Jha <mayank25080562 at gmail.com> wrote:

> However unfortunately whenever it came to Hindi my choices got narrowed to 3
> or 4 fonts only as compared to at least 40-50 choices for the Roman script.
> Thus, making this tool would allow font developers to contribute more good
> quality Open Type fonts for Indic script.

One aspect we would like to keep in mind is while the GSoC cycle might
limit us to creating a set of tests/rules for a particular script, the
framework devised should make it easy to extend the coverage to other
script. A tool-service like this is useful not only for scripts for
Indian languages, but world-wide.

> Implementation Details:
>
>
> I will be implementing this application for the Devanagri script. To
> implement the testing application what I propose to do is, first of all set
> standards or rule-sets ie what to search for in any Devanagri script. There
> exists a current application called fontQA but that runs only inside
> FontLabs, which is not freely available, therefore it is very difficult to
> try, modify and test it.

I would recommend revisiting the goal. The original idea was to have a
framework/test-harness and, use a script to demonstrate the
proof-of-concept. So, in effect, the pressing need is to think over
the architecture of the framework, how to make writing the tests easy,
how to handle grouping of tests across various elements and so forth.

> Illustrating what I would like to look for in the Devanagri script, ie the
> standards followed by uniscribe used by microsoft . Is listed below:

The OpenType specification (as implemented by Uniscribe and Harfbuzz)
are not the only guidelines related to fonts. W3C provides
recommendations as well. I am bit disappointed that you have not
demonstrated the initiative to refer to them.

> Positioning of dots, dot is present at the bottom of the consonant.
>
> How joining of conjoined consonants is done like the letter ksha.
>
> Positioning of reph over the base line when the letter ra is conjoined with
> any consonant.
>
> Below base conjoining of ra with a consonant, like dra.
>
> Half form of consonants, wherein only half form of a consonant is used, like
> the ka in the word waqt.
>
> Conjunction of ta with ra .
>
> Conjunction of da as in the word vidya.
>
> The matra of e used in consonants.
>
> Formation of reph.
>
> Use of anuswar and chandra bindu with other vowel matras.
>
> Use of halant.

Are the above points listed by you specific to the script or, do they
have elements of generality considering other scripts, especially
those for Indian languages? And, how do you propose to check for
slant? Additionally, what about font weights and point sizes?

> The above points can be extended further, to improve quality assessment.
>
>
> These characteristic features would be identified in some standard font,
> like MANGAL or which is the best available standard. Two fonts will be
> compared based on how close/similar they are to these standards, and the
> developer would be shown the shortcomings in each of the form so that they
> will have point wise difference between the fonts, which otherwise needs
> qualitative judgement.

It would probably be a good idea to avoid getting into preferences of
one font over another. Consider certification marks provided by bodies
like BIS for instance - the unit-under-test is pitted against a
specification and, the inferences drawn. There is no requirement to
have a baseline reference implementation.

> To really compare two fonts we would be needing to compare their GSUB and
> GPOS methods used when formation of the text is done. So I would be using
> the FreeType library to analyse the fonts.

I am not too sure what you mean when you use the phrase "to really
compare two fonts". To me it appears that you infer that the process
of qualitative assessment is not "really comparing" anything. If that
is so, it is a somewhat correct conclusion. What the proposed idea
intends to do is provide a method to convert well-accepted and used
guidelines into rules-for-testing and, make it easy for people to
extend the framework or, contribute/re-write tests.

Additionally, when you are looking at fonts, please do keep in mind
their licenses as well. It would be good (for the proof-of-concept) to
use fonts that have non-restrictive licenses.

--
sankarshan mukhopadhyay
<https://twitter.com/#!/sankarshan>