Root lists.

[Koontz John E]

On Tue, 15 Oct 2002, Linda A Cumberland wrote:
> But thanks for the algorithm, John.  I'll show it to Wally Hooper (at
> the Institute) and see what he thinks.

Wally's going to laugh at my calling it an algorithm.  It's incredibly
simple.

The idea is that you have a database which, in simple terms, looks like
this:

hw bazi'be
def pinch
ref page 2

hw zaN'de
def grive
ref page 4

hw a'baz^ibe
def raise tent
ref page 10

and you search hw fields for patterns like

^(wa|we|a|u|i|)(ga|ba|bi|dha|dhi|mu|ma|na|naN)

(i.e., a start of form followed by any of wa, we, etc., or nothing
followed by any of ga, ba, etc.  You might want to add possessive, etc.,
forms, e.g., insert {gi|g|kki|) into the list between the other two
sublists, etc.  You have to allow for e from agi, and some other
complications I haven't allowed for, but this is the general idea.

When you match a form with this pattern you've got the part it matched
and (with a bit more added to the pattern) also the part not matched, and
you print the two in reverse order with a comma and a space between them
and the defintion and reference appended, getting, e.g.,

zibe, ba- pinch page 2
z^ibe, a'ba- raise tent page 10

You sort the output on the first field and you're there.  Imagine many
more examples, of course, to make it interesting.

I've neglected accent, but you can handle that by deleting accent marks
before doing the match or by allowing 1 or no accent marks after each list
item, etc.  The second approach is probably better, since you have the
accents to reinsert.

Note that one can cross compare several languages by inserting in from of
each line in the output a "back reconsruction" of the supposed root.

For example, replace all b with p and all final e with -, yielding

zip- zibe, ba- pinch page 2
z^ip- z^ibe, a'ba- raise tent page 10

Then sort again.  Here I'm back transforming Omaha, but if you had an IO
file with

dhiwe, gi- whittle ...
ghiwe, ba- gouge ....

that would bve processed by substituting p for w, z for dh and - for final
e to yield

zip- dhiwe, ga- whittle ...
ghip- ghiwe, ba- gouge ...

These could be sorted together with lists from other languages because the
sort keys (the first fields) are commensurate.

You can get a lot out of simple operations like match, extract, transform,
and sort.  Of course, you need the databases to start with, and there's
where the Siouan Archives come in, as well as other files assembled by
others.  Also, you need a tool that does operations like this.  At present
the easiest are scripting languages like AWK, TCL, Perl, and so on.
SNOBOL or SPITBOL would work, too, or the internal scripting languages
built into some text editors, like emacs and its relatives.  I say
easiest, but some of these may not strike some folks as easy.

JEK