Resource: STO

Reference STO SprogTeknologisk Ordbase (Danish Lexicon for NLP/HLT Applications)
Date of Submission Jan. 24, 2014, 4:31 p.m.
Status accepted
ISLRN 050-677-531-676-8
Resource Type Lexicon
Media Type Text
Source
Language Danish
Description

The STO Lexicon is the most comprehensive computational lexicon of Danish comprising approx. 81,530 entry words, and it is well integrated with the European activities in the field of lexicon development building on experience obtained from the PAROLE and SIMPLE projects. The model and descriptive method of the STO lexicon are kept compatible with the architecture and descriptive language of PAROLE/SIMPLE. A number of refinements, adaptations and language-specific extensions to the basic model are implemented in STO.

Lexical coverage and encoded information by category is distributed as follows:

Lexical Category Lemmas
Morphology
only Morphology + Syntax Morphology + Syntax + Semantics
Noun 64,735 47% 41% 12%
Adjective 9,773 32% 55% 13%
Verb 5,775 2% 81% 17%
Adverb 771 81% 0% 0%
Interjection 158 100% 0% 0%
Preposition 80 100% 0% 0%
Conjunction 60 100% 0% 0%
Pronoun 445 100% 0% 0%
Misc.: Determiner, adposition, conjunction, etc. 128 100% 0% 0%
Total 81,524

A part of this vocabulary (i.e. 12,060 lemmas) is selected from 6 domains, as follows:

Domain Nouns Verbs Adjectives Total
IT 1,957 52 66 2,075
Environment 2,055 48 285 2,388
Commerce 1,537 16 57 1,610
Administration 2,435 25 193 2,653
Health 1,603 42 350 1,995
Finance 1,258 24 57 1,339
Total 10,845 207 1,008 12,060

Linguistic coverage / Main information types:
Morphology: PoS, inflectional patterns, agreement features, noun compounding, spelling variants, etc.
Syntax: subcategorisation frames (categorical and functional valency), alternation, diathesis, reflexivity, etc.
Semantics: encoded at three different levels of specificity. The most specific is Level 3, which contains sense distinction, ontological type, argument structure, selectional restrictions, qualia structure, event structure, domain information, etc. Level 2 is a proper subset of Level 3 representing a more lean semantics (without qualia and event structure, etc.) whereas Level 1 concerns information on source domain only.

The resource was validated internally. This lexicon is well suited for NLP/HLT monolingual applications, as lexicon component in taggers, parsers, grammar & spell checkers, summarisation tools, web crawlers, computer-aided language learning, as well as multilingual applications; also possibility for linking to other PAROLE/SIMPLE-compatible resources.

The lexicon is provided with a thorough documentation in English and distributed on CD-ROM.

Version 1.0
Distributor ELRA