Korp Karp Installation

The Korp code:

• Korp Backend
• Korp Frontend
• Corpus pipeline

Links to the Karp code are forthcoming.

• Download Korp code
• Install at gtweb
• Install corpora
• Make interface

• Free
  • skuvlahistorja1-6
  • fad
• Bound
  • news
  • ficti
  • NT

• smesme: news + ficti
• nob2sme: fad + skuvlahistorja1-6
• smedep: news + ficti + facta/skuvlahistorja1-6 + bibel/newtestament

Menu:

1. search for sme wordforms (kwic-snt in corpus ccat) – corpus: smesme
2. search for sme lemmas (kwic-snt? in analysed corpus syn) – corpus choices: smesme, nob2sme
3. search for sme and nob in translations (lemma search in sentence aligned sentences) – corpus: nob2sme
4. deepdict sme (lemma search -> dependency daughters in corpus dep) – corpus: smedep

Definitions

• lexeme = member of an open lexical category, having meaning and form but being neither
• lemma = wordform used as representative for lexeme
• grammatical word pair of lemma+grammatical properties and wordform
• paradigm = set of grammatical words realising a lemma
• lemgram = set of wordforms in paradigm

Generation

Generation of lemgrams from lexc:

Use dict-isme-norm.fst or generator-dict-gt-norm.xfst or generator-dict-gt-norm.hfst. We remove the tags v1, v2.. from the fst. It is better for the user that all variants of the same paradigm are in the same lemgram. Many fst-lemmas have more than one entry in lexc, so the list should be uniqed before generating forms. I suggest that we start with these files:

For nouns, we pick different 3 lists: The ordinary nouns, the actors (NomAg), and the G3-marked nouns. For the other parts of speech, one command is enough. Commands to filter (ir)relevant forms:

noun-sme-lex.txt:

• Ordinary words:

egrep -v "(G3|ACTOR|CmpN/Only|ShCmp|RCmpnd|\+V\+|^\!)"

• ACTOR:

grep N+NomAg

• G3:

grep N+G3

verb-sme-lex.txt:

egrep -v "(ENDLEX|\+V|^\!)"

adj-sme-lex.txt:

egrep -v "(LEXICON|Der| Rreal | R |^\!)"

adv-sme-lex.txt:

egrep -v "(LEXICON| K |^\!)"

• 2013: 9.4.   // 4.12.
• 2014: 8.1..