N-grams from the National Library of Norway 2026

This resource contains n-grams – i.e. uni-, bi- and trigrams – from all books and newspapers that had been digitized at the National Library of Norway as of March 2026. The n-grams have been extracted from material consisting of approximately 685,000 books and 4,720,000 newspapers, amounting to a total of 160 billion tokens (words and punctuation). The file format is UTF-8-encoded JSONL, compressed with the efficient Zstandard algorithm.

Content Keys in the n-gram JSONL files: – first – the first word (in uni-, bi- and trigrams) – second – the second word (in bi- and trigrams) – third – the third word (in trigrams) – lang – the language of the n-gram (only for books, the newspapers have no language classification as yet) – freq – the total frequency of the n-gram in the collection of books and newspapers – counts – a dictionary with raw frequencies for each year totals.json contains aggregated frequencies per year in the book and newspaper corpora. Using them, relative frequencies can be calculated in order to compare frequencies over time, as in NB N-gram. ngram-2026-digibok-metadata.jsonl and ngram-2026-digavis-metadata.jsonl contain simple metadata for the books and newspapers. More extensive metadata can be obtained through Oria or the APIs at https://api.nb.no/.

Data preparation The n-grams were extracted from the full-text databases provided by the DH-lab at the National Library of Norway (https://www.nb.no/dh-lab/). The following frequency cuts were applied:

• unigram (books): total frequency of less than five in the corpus for Norwegian Bokmål, less than two in all other languages. Additionally, the n-gram must appear in more than one year.
• unigram (newspapers): total frequency of less than five in the corpus. Additionally, the n-gram must appear in more than one year.
• bigram (books and newspapers): similar frequency cuts as for unigrams. Additionally, hapax bigrams within one year were removed.
• trigram (books and newspapers): similar frequency cuts as for bigrams. Additionally, hapax trigrams within one year were removed.

License The data are released in the public domain (CC-0).