SV: [Snowball-discuss] Norwegian stemmer

[Bruusgaard, Jan]

Martin,

Here are my comments. As I see it, there is an improvement for 13 words
and 5 words are worse. 

In all the 5 cases the s is a genitive s. I would also belive that many
of the 13 words have a higher frequency in normal text.


-ks word                      current action                Bruusgaard
adjustment
--------                      --------------
---------------------

boks                          bok                           + boks
brokks                        brokk                         0 brokks
danmarks                      danmark                       0 danmarks
fisks                         fisk                          0 fisks
foretaks                      foretak                       0 foretaks
heks                          hek                           + heks
innenriks                     innenrik                      + innenriks
inneriks                      innerik                       ? inneriks
(spelling error innenriks +)
instruks                      instruk                       + instruks
juks                          juk                           + juks
laks                          lak                           + laks
markedsinndeks                markedsinndek                 +
markedsinndeks (spelling error markedsindeks)
paradoks                      paradok                       + paradoks
seks                          sek                           + seks
straks                        strak                         + straks
styreinstruks                 styreinstruk                  +
styreinstruks
utenriks                      utenrik                       + utenriks
verks                         verk                          0 verks
voks                          vok                           + voks
 

Jan


-----Opprinnelig melding-----
Fra: Martin Porter [mailto:martin.porter at grapeshot.co.uk] 
Sendt: 20. april 2005 11:52
Til: Bruusgaard, Jan
Kopi: snowball-discuss at lists.tartarus.org
Emne: Re: [Snowball-discuss] Norwegian stemmer



Jan,

I apologise for taking so long to get back to the stemming work, but I
have been busy with other things. You suggested (13 Jan) removing k from
define_s_ending in the Norwegian stemmer. In the sample vocabulary about
one word per thousand is affected by this. Here is the list:

-ks word                      current action                Bruusgaard
adjustment
--------                      --------------
---------------------

boks                          bok                           + boks
brokks                        brokk                         0 brokks
danmarks                      danmark                       0 danmarks
fisks                         fisk                          ? fisks
foretaks                      foretak                       0 foretaks
heks                          hek                           + heks
innenriks                     innenrik                      ? innenriks
inneriks                      innerik                       ? inneriks
instruks                      instruk                       + instruks
juks                          juk                           + juks
laks                          lak                           + laks
markedsinndeks                markedsinndek                 ?
markedsinndeks
paradoks                      paradok                       + paradoks
seks                          sek                           ? seks
straks                        strak                         ? straks
styreinstruks                 styreinstruk                  ?
styreinstruks
utenriks                      utenrik                       + utenriks
verks                         verk                          ? verks
voks                          vok                           + voks

I think I will add your suggestion, since the forms marked + (I believe)
are improved by the rule, the forms marked 0 made worse. The forms
marked ? I cannot interpret.

I wonder if you could comment on this list as a native speaker of
Norwegian?

Martin