Re: Automatic clustering of languages

From: Richard Wordingham
Message: 48173
Date: 2007-04-02

--- In cybalist@yahoogroups.com, "mkelkar2003" <swatimkelkar@...> wrote:

> Yes, but the 'others' group splits into 'Dravidian' and then after
> another split into Greek and Lithuanian/Lativian the, latter being
> very close to Sanskrit.

The 'others' group contains *all* the non-Indo-European languages in
the study.

I suspect that the Graeco-Baltic group is largely the group where
masculine substantives have final -s in the nominative singular. The
study comes close to isolating a Finno-Maori group. The latter is
probably capturing a similarity between typical Finnish words and the
typical Maori words. Remember that the metrics are quite crude, and
they do not capture the notion of a regular sound change. It would be
fun to obtain the analysis program and see how sensitive the analysis
was to minor changes in the presentation of the data. I don't have
time to do it.

Richard.