🚀 go-pugleaf

Forwarded message:
>From jaap@research.att.com Wed Oct 20 18:59:30 1993
Message-Id: <m0ophec-0004AlC@hunny.research.att.com>
To: jbowyer@cis.vutbr.cz
Subject: Re: CA: Hyphenation
Return-Receipt-To: jaap@research.att.com
In-Reply-To: Your message of Wed, 20 Oct 93 17:42:04 +0100.
             <9310201642.AA07787@rhino.cis.vutbr.cz>
Date: Wed, 20 Oct 93 13:48:57 D
From: Jaap Akkerhuis <jaap@research.att.com>


In the consolidated answers to the original question from Larry
Spitz (spitz@pal.xerox.com):

	 I am interested in rules for hypenation in European
	 languages and in particular how they differ from language
	 to language. Any pointers would be appreciated.

Lode Goukens (@earn.cvut.cz:PS890866@FPSW.UFSIA.AC.BE) writes:

    Most Dutch newspapers use Italian hyphenation systems because
    of the problems.

I don't think that that's true, at least in the Netherlands. There
might be some Dutch newspaper who do this, but I doubt it. Most
Dutch newspaper use a more intelligent algorithm.  Often this is
based of ``Beste Splits'', developed in the years 60 at the
Mathematical Center (Now called CWI; Centre for Wiskunde &
Informatica). It uses some statistics to guess the best hyphenation
point. This method has been solded to various newspaper typesetting
systems vendors. Of course a lot of them made improvements since
then.

The original algorithm does a fair job of German & Swedish as well,
(at least, that is what I've been told after I put it as an optional
algorithm in troff. At least, much better then the original
(American-) English one).

    In Dutch hyphenation is rather simple at first sight, but some
    words contain parts that are very common but yet different.
    Eg. in Dutch two words can be put together to form a new one.
    But when the software doesn't see this it is possible that
    during hyphenation it splits the second 'word' because of the
    rule that sais to split after the first non-vocal if possible.

``Beste splits'' did/does a fair job for compound words.  An obvious
improvement, which we didn't do, is to hold the to be hyphenated
word against a dictionary to split it in the components in separate
words (which is of course the ideal hyphenation point), and then,
when necessary hyphenate the individual components.

A major problem (with a lot of languages) are homonyms (one or two
words that have the same sound and often the same spelling but
differ in meaning). Often the hyphenation is different for different
meanings, so the best hyphenation point then becomes context
sensitive.

	jaap

PS. I get the impression that Lode is from Belgium. It might be
that his claim is true for Flemish papers published there.