tekstikorpus

Tekstikorpus, commonly referred to in English as a text corpus, is a digitally stored collection of authentic written texts used for linguistic research and natural language processing. A Tekstikorpus is designed to be representative of a language or domain and is annotated with metadata such as source, date, author, genre, and licensing information. Texts are typically processed by tokenization and normalization to create a machine-readable resource suitable for quantitative analysis.

Corpora vary widely in scope and structure. General corpora aim to cover broad language use, while specialized corpora focus on particular genres, domains, or time periods. They can be monolingual or multilingual, balanced or unbalanced, raw or annotated. Common annotations include part-of-speech tags, lemmas, morphological features, syntactic parses, and semantic roles. Corpora may also include word frequency lists and collocation data.

The collection and design of a Tekstikorpus involve sampling decisions, data sourcing, copyright considerations, and ongoing maintenance. Access ranges from open, freely downloadable data to restricted, license-based use, often with documentation on data provenance and quality.

Applications include linguistic research, lexicography, and the development of natural language processing tools such as language models, search engines, and grammar checkers. Tekstikorpus resources underpin statistical analyses, corpus-based learning, and empirical studies of language use over time.

See also: corpus linguistics, language resource, annotated corpus, frequency analysis, natural language processing.