Grammatikkdataene

Grammatikkdataene is a term used in Norwegian to describe structured collections of grammatical information about a language. They are designed to capture how words change form, how phrases are built, and how sentences are organized, enabling systematic analysis and computational processing.

A typical grammar data set includes morphological annotations (such as part-of-speech tags, lemmas, and features like gender, number, case, tense, and mood) and syntactic information (such as dependency or constituency structures). Some datasets also provide inflection paradigms, multiword expressions, and phonological or orthographic notes. Data may be aligned with established annotation schemes to facilitate cross-project comparability.

Sources and preparation methods vary. Grammatikkdataene are often derived from annotated corpora, grammars, and dictionaries, sometimes supplemented by parallel corpora for translation research. Annotations are usually produced by linguists, with quality assurance through detailed guidelines, training, and measures like inter-annotator agreement. Semi-automatic tagging with human validation is common, especially for large datasets.

Uses and applications are broad. They support natural language processing tasks such as part-of-speech tagging, syntactic parsing, grammar checking, and machine translation, as well as linguistic research and language education tools. Researchers also use grammar data to study typology, language variation, and historical linguistics.

Access and standards vary by project. Some Grammatikkdataene are open data under permissive licenses, while others are restricted to researchers or institutions. Initiatives like Universal Dependencies provide widely adopted reference models for cross-language grammatical annotation, influencing many Norwegian and other-language datasets.