Re: New member questions

--===============8991425086159734330==
Content-Type: multipart/alternative; boundary="000000000000210526062c7a1715"

--000000000000210526062c7a1715
Content-Type: text/plain; charset="UTF-8"

Harri,

You will have to excuse me for not being completely on top of the subject
of snowball these days, but the main work was done a quarter of a century
ago, and I am now 80 years of age. Papers evaluating snowball were often a
bit negative: that is because they were testing some new system of the
authors with a base system, and the base system would often use snowball.
If they outperformed the base system their paper would be published, if not
they would hold it back and try again. There was a general paper with
comparisons of IR with/without snowball for several languages, which put
snowball in a much better light, but I can't quite find it at the moment.
If I do I'll let you know. But for Finnish itself this is encouraging,

http://clef.isti.cnr.it/2004/working_notes/WorkingNotes2004/16.pdf

with their conclusion,

"For Finnish, the performance is quite high. The snowball stemmer works
very well."

(The stemmers do not reduce words to real vocabulary words incidentally,
just to a character string that collects variant forms together.)

I think the final statement on these stemming algorithms must be that they
are a simple and inexpensive way of conflating variant forms of a word
together, and that this can be useful in certain circumstances.

Martin

--000000000000210526062c7a1715
Content-Type: text/html; charset="UTF-8"
Content-Transfer-Encoding: quoted-printable

<div dir=3D"ltr">Harri,<br><br>You will have to excuse me for not being com=
pletely on top of the subject of snowball these days, but the main work was=
 done a quarter of a century ago, and I am now 80 years of age. Papers eval=
uating snowball were often a bit negative: that is because they were testin=
g some new system of the authors with a base system, and the base system wo=
uld often use snowball. If they outperformed the base system their paper wo=
uld be published, if not they would hold it back and try again. There was a=
 general paper with comparisons of IR with/without snowball for several lan=
guages, which put snowball in a much better light, but I can&#39;t quite fi=
nd it at the moment. If I do I&#39;ll let you know. But for Finnish itself =
this is encouraging, =C2=A0<br><br><a href=3D"http://clef.isti.cnr.it/2004/=
working_notes/WorkingNotes2004/16.pdf">http://clef.isti.cnr.it/2004/working=
_notes/WorkingNotes2004/16.pdf</a><br><br>with their conclusion,<br><br>&qu=
ot;For Finnish, the performance is quite high. The snowball stemmer works v=
ery well.&quot;<br><br>(The stemmers do not reduce words to real vocabulary=
 words incidentally, just to a character string that collects variant forms=
 together.) <br><br>I think the final statement on these stemming algorithm=
s must be that they are a simple and inexpensive way of conflating variant =
forms of a word together, and that this can be useful in certain circumstan=
ces.<br><br>Martin<br><br><br></div>

--000000000000210526062c7a1715--


--===============8991425086159734330==
Content-Type: text/plain; charset="us-ascii"
MIME-Version: 1.0
Content-Transfer-Encoding: 7bit
Content-Disposition: inline

_______________________________________________
Snowball-discuss mailing list
[email protected]
https://lists.tartarus.org/mailman/listinfo/snowball-discuss

--===============8991425086159734330==--