[R] Categorizing Fonts using Statistical Methods

[Leonard Mada]

Hello Johannes,


Johannes Hüsing wrote:
> Leonard Mada <lmada_at_gmx.net> [Sun, May 04, 2008 at 07:26:04PM CEST]:
> > Dear list members,
> >
> > Every "modern" OS comes with dozens of useless fonts, so that the
> > current font drop-down list in most programs is overcrowded with fonts
> > one never will use. Selecting a useful font becomes a nightmare.
> >
> > In an attempt to ease the selection of useful fonts, I began looking
> > into sorting fonts using some statistical techniques. I summed my ideas
> > on the OpenOffice.org wiki:
> > 
> http://wiki.services.openoffice.org/wiki/User_Experience/ToDo/Product/Font_Categories
> >
> > Of course, there is NO guarantee that something useful will emerge, but
> > at least someone has tried it.
> >
>
> Why is there nothing mentioned with respect to the classical font 
> categorization, Venetian, Aldine, Transitional, Modern, Slab Serif, ... ?

I played with the idea over and over again, but decided then against it.

I had a look both on the Adobe site, and on various other sites (e.g. 
http://graphicdesign.spokanefalls.edu/tutorials/process/type_basics/type_families.htm#oldstyle).Unfortunately, 
fonts belonging to different families may look very similar, while fonts 
within one family are different enough to warrant a distinct 
classification. Especially this latter aspect makes me think that the 
font families are not that helpful, and - when choosing the appropriate 
font -  I do NOT want to limit myself to one family. A different font 
family might look even better.

Also, I cannot remember a  single time I have used a font based on its 
family. Rather, a font gets selected based on how it looks within a 
specific document (well, mostly it gets selected because the person 
knows it - but lets ignore this and adopt a more scientific approach).

Selecting some measures, like font width, height, weight, complexity, 
compactness, slant, [...] seems a sensible approach.

> [...]
> > - maybe some other measures
>
> If you can obtain the *.afm information of the font, you have some 
> useful parameters such as cap height, ascender height, descender 
> height, oblique angle ...

I do have a rather limited understanding of the font-files proper. If I 
am correct, .afm-files are available only for post-script fonts. Of 
course, on Windows, most fonts will be TrueType and OpenType. I have no 
idea, IF such information is available for these fonts.

My primary problem is however, that the purpose of this analysis is to 
let end-users perform this same analysis on their computers on their own 
font sets. My plan was to do a proof of concept analysis in R, and later 
(when I have some better idea how to categorise fonts and everything 
works fine) to post such a feature request in specific programs.

At this point, this sorting of fonts is of unproven benefit and of 
unknown behaviour. So, I wouldn't want to waste developers time into 
something that might prove useless (though I have high expectations that 
something useful will emerge - NOT sure however which of the specific 
measures will bring the most differentiating features).

I still hope in completing succefully this task.

Many thanks for your advice, I will take another look at afm-files.

Sincerely,

Leonard