GiellaLT

GiellaLT provides an infrastructure for rule-based language technology aimed at minority and indigenous languages, and streamlines building anything from keyboards to speech technology.

View GiellaLT on GitHub

Page Content

How to handle variation in lexc

Non-normative forms: Err/Orth

Non-normative variant with inflection

ex.: “fálahas” is a non-normative variant of the lemma “fáláhas”, and it inflects. The normative form on the left side, and so the lemma in the analysis will be a normative form and can be found e.g. in the dictionary.

fáláhas:fáláhass JOHTOLAT ;
fáláhas+Err/Orth:fálahass JOHTOLAT ;

The descriptive FST will inflect both fálahas and fáláhas, but the line with the tag Err/Orth is removed from the normative analyser/generator during the compilation prosess.

fáláhasat 
fáláhasat	fáláhas+N+Pl+Nom

fálahasat
fálahasat	fáláhas+Err/Orth+N+Pl+Nom

The normative analyser:

fáláhasat 
fáláhasat	fáláhas+N+Pl+Nom

fálahasat
fálahasat	fálahasat	+?

Non-normative variant without inflection (hard coding)

ex.: “fálahas” is a non-normative variant of the form “fáláhas”, and it does not inflect, and therefore it does not get a continuation lexicon with inflection for nouns.

The normative form on the left side, and so the lemma in the analysis will be a normative form and can be found e.g. in the dictionary.

fáláhas:fáláhass JOHTOLAT ;
fáláhas+N+Sg+Nom+Err/Orth:fálahas ENDLEX ;

The word itself is non-normative: Err/Lex

Ex. brillefutterála which is a slightly adapated loanword from Norwegian to North Saami. The normative word is čalbmelássaskuohppu

brillefutterála+Err/Lex:brille#futterál SOSIAL

The descriptive FST will inflect brillefutterála, but the line with the tag Err/Lex is removed from the normative analyser/generator during the compilation prosess.

brillefutterálat 
brillefutterálat	brillefutterála+N+Pl+Nom

The normative analyser:

brillefutterálat
brillefutterálat	brillefutterálat	+?

Lexical homonymi: how to identify the correct lemma e.g. in a dictionary

Two lemmas, which base forms are homonyms, have different paradigms and semantics.

The lemmas belong to different stem-categories: Add morphogical tags

Example from North Saami. G3 tag for Grade 3 for consonantgradation with geminate in lemma, e.g. ss:

beassi:beassi BEARRI "reir" ;
beassi+G3:beas'si AIGI "never" ;

Analysis:

beassi
beassi	beassi+N+G3+Sg+Nom
beassi	beassi+N+G3+Sg+Acc
beassi	beassi+N+G3+Sg+Gen
beassi	beassi+N+Sg+Nom

beasi
beasi	beassi+N+Sg+Gen
beasi	beassi+N+Sg+Acc

Example from North Saami. NomAg tag for derivation Nomen Agent

vuovdi+NomAg:vuovdi ACTOR "salesman" ;
vuovdi:vuov'di AIGI "forest" ;

Analysis:

vuovdi
vuovdi	vuovdi+N+NomAg+Sg+Nom
vuovdi	vuovdi+N+NomAg+Sg+Acc
vuovdi	vuovdi+N+NomAg+Sg+Gen
vuovdi	vuovdi+N+Sg+Nom

vuovddi
vuovddi	vuovdi+N+Sg+Gen
vuovddi	vuovdi+N+Sg+Acc

There is no reason for adding morphogical tags: Add homonymi tags

Example from South Saami:

govledh+Hom1:govl TJOEHPEDH_TV "höra" ;
govledh+Hom2:govl VÅÅJNEDH "höras" ;

Analysis:

gåvla
gåvla	govledh+Hom1+V+TV+Ind+Prs+Sg3

govloe
govloe	govledh+Hom2+V+IV+Ind+Prs+Sg3

Orthograpic variants of the same lemma: tags v1, v2…

Orthograpic variants of the same lemma, for base form and at least parts of the inflection paradigm, should be under the same lemma. But we can add a variants tag as a help to recognize the correct base form for the paradigm.

Example from North Saami.

mandáhta+v2:mandáhtta GOAHTI-A ;
mandáhta+v1:mandáhta STAHTA ;

If the base forms are identical, but there are variants in the inflection, we don’t use these tags.