140116

• web page
  • Online analyser
  • Form generator
  • Paradigm generator
• pikiskwewina
  • as dictionary and click-in-text
• Spellchecker
  • for the same text
• learning tools based on Saami slides

syllabics conversion

Goal:

• to have two spellcheckers (or one spellchecker understanding both)
• Romanian as Roman
• Bulgarian as Syllabics
• Syllabics conversion online
• demo syllabics

TODO

Trond to:

• Send a list of relevant Oahpa presentations (Lene)
• Look at cgi-bin: conversion to/from syllabics
• look at the wrapped syllabic fst

Sjur to:

• look at the wrapped syllabic fst
• Take a trip to Bulgaria

Conor to:

• Add words, look at the linguistics

How to have a speller

go to langs/crk, and issue these commands:

./configure --with-hfst --enable-spellers
make
sudo make install

• Download LibreOffice if you do not have it already
  (must be version 4.1 or newer).
• download & install oxt (Mac version):
  http://divvun.no/static_files/voikko-2013-06-26.oxt
  • restart LibreOffice if it was already open
• Open a new text document, write some text, select all
  • e.g. in the lower margin of your document it says: Default Style US English
• Click on the language name, and choose More...
• set the language to Romanian (or soon: Bulgarian). Romanian (and Bulgarian) should have a blue "ABC" in front of it in the language drop down menu).

To update the speller:

• add a new noun
• make
• sudo make install
• restart LibreOffice

To repeat this on Windows:

• Download and install LibreOffice (at least 4.1)
• download and install the Windows beta oxt: http://www.puimula.org/htp/testing/voikko-sma-fi.oxt
• build the speller on a mac or linux machine, final file can be found in: crk/tools/spellcheckers/fstbased/hfst/crk.zhfst
• copy this zhfst file to the Windows machine, place it in: to-be-added

Text for the presentation

http://www.ualberta.ca/~arppe/PlainsCree.html

The text:

nitêminân nipâw sisone iskwatemihk.
dog+N+Pl1Ex sleeps+V beside+PART door+N+LOC

waniskâw kîkisepâ.
wakes.up+V morning+N 

waniskâw ekwa nohtekatew.
wakes.up+V and+PART is.hungry+V

wâpahtam ôskanisis wiyâkanihk ekwa mîciw.
sees+V bone+N+DIM bowl+N+LOC and+PART eats+V

ekota-ohci nôhkwâtam wiyâkan.
here.from+PART licks+V bowl+N

keyâpic nohtekatew ekwa kâwe nipâw.
still+PART is.hungry+V and+PART again+PART sleeps+V

Issues to fix:

• waniskâw
  • Trond to fix the waniskâw x 14 (due to Trond's flag diacritic experiments).
• nohtekatew nohtekatew +?
  • VAI, to be added
• nôhkwâtam nôhkwâtam +? Conor: Check with Dorothy
  • VTA ??
• mîciw mîciw +?
  • VTA to be added
• ekota-ohci ekota-ohci +?

Working with a productive analyser for now.

Conor has checked files and added verb types. Different verb types are to be checked in.

Subjunctive prefix ê-

In writing:

• Either: always ê-
• Or: êh in front of vowels, no mark (thus, just an ê) in front of consonants

Solution:

Two prefixes in lexc:

• ê-
• êh^%eh

and then to h deletion before %^eh in twolc

The tag is now +Sbj, an alternative is +Conjunct mode, +Cnj, so we could do that.

Decided:Use +Cnj

More words in the dictionary, especially the words of the text. Input here: comma separated stuff to crkeng/inc/:

main/words/langs/crkeng/inc/nouns.csv

cat inc/nouns.csv 
atim	n	dog
inini	n	man	
nâpês	n	boy	
apiscacihkos	n	antelope
mâyatihk	n	bighorn sheep
atihk	n	caribou
apisimôsnos	n	deer
wâwaskêsniw	n	elk
mistatim	n	horse
môswa	n	moose
maskwa	n	bear
okistatowân	n	grizzly bear
wâpask	n	polar bear
sîsîp	n	duck
môhkomân	n	knife
sakâw	n	soup
mîcimâpo	n	soup
sîsîpâwi	n	duck egg
wâwi	n	egg
iskwêw	n	woman
nipâw	v	sleep
wâpam	v	see

  <e>
    <lg>
      <l pos="N">sakâw</l>
    </lg>
    <mg>
      <tg>
        <t pos="N">soup</t>
      </tg>
    </mg>
  </e>

Results to be added to

main/words/langs/crkeng/src/N_crkeng.xml

Procedure for updating the dictionary (needed: account on the gtweb machine):

/dicts/nds/NDSUpdatingDictionaries.html

(with pikiskwewina or guusaaw as the variable for DICT)

1. Log in to the server via SSH as the NDS user
   in the following way:
   ssh neahtta@gtweb.uit.no
   with a password
2. Then follow the instructions on the page

Many writers do not write ê. The analyser handles both, but we should consistently always write e.g. with macron in the code.