The NKJP_1M corpus

The NKJP_1M corpus is a manually annotated one million word subcorpus of the National Corpus of Polish (NKJP – Narodowy Korpus Języka Polskiego), composed of various text samples (see below). It is a corpus of contemporary Polish with texts published after the year 1945; it contains written, spoken and web communication. The corpus features lemmatisation, morphological annotation, and representative coverage of text categories.

Name NKJP_1M
Positions Number of positions (tokens) 1,215,513
Number of positions (excl. punctuation) 992,014
Number of word forms 143,477
Number of lemmas 54,174
Structures Number of documents <doc> 3,889
Number of paragraphs <p> 18,484
Number of sentences <s> 85,663
Further information Reference corpus YES
Representative corpus YES
Publication year 2018

Text classification

The classification of NKJP_1M texts combines traditional criteria with a categorisation based on topic and genre. Genre categories (or types, in Polish corpus terminology) are often integrated with medium categories (or communication channels, in Polish corpus terminology), but formally, these two types of categorisation remain separate.

Communication layer doc.genre Category Proportion
written #typ_publ journalism 48.85%
#typ_lit fiction 17.04%
#typ_fakt non-fiction 5.34%
#typ_inf-por informative texts 5.62%
#typ_urzed legal texts 2.97%
#typ_nd popular science texts 1.91%
#typ_nklas non-fiction unclassified book 1.00%
#typ_listy correspondence 0.04%
#typ_lit_poezja poetry 0.01%
spoken #typ_qmow quasi-spoken texts 2.50%
#typ_media spoken media text 2.07%
#typ_konwers spoken conversational texts 5.57%
web #typ_net_interakt interaction-based Internet texts 5.18%
#typ_net_nieinterakt non-interaction-based Internet texts 1.91%

Positional annotation and tagging

Compared to typical corpora of Czech, NKJP_1M additionally has a positional attribute which is specific for Polish, the so-called flexeme. It is a category which further subdivides parts of speech into more specific lexeme classes. For instance, within nouns (subst), depreciative nouns (depr) form one of the flexeme subgroups; flexemes also distinguish between regular adjectives (adj), the first part of compound adjectives (adja, e.g. biało-czerwony, sportowo-rekreacyjny), post-prepositional adjectives (adjp, e.g. po polsku, od dawna), and predicative adjectives (adjc, e.g. jestem pewien, był wesół i zdrów); and there is a particularly fine-grained subcategorization of verbs (more than 10 different flexemes).

Moreover, the Polish tagset differs from the Czech one; its detailed description (including the full flexeme list) is available here.

In addtion, two positional attributes were added to the original corpus: lc and lemma_lc, which allow to search the corpus in a case-insensitive manner.

How to cite NKJP_1M

Przepiórkowski,. A. – Degórski, Ł. – Murzynowski, G. – Szałkiewicz, Ł. – Czelakowska, A.– Savary, A. – Głowińska, K.: NKJP_1M: ręcznie znakowany milionowy podkorpus NKJP. Institute of the Czech National Corpus, Faculty of Arts, Charles University, Prague 2018. Available at WWW: http://www.korpus.cz

Degórski, Ł. – Przepiórkowski, A. (2012): Ręcznie znakowany milionowy podkorpus NKJP. In: A. Przepiórkowski – M. Bańko – R. L. Górski – B. Lewandowska-Tomaszczyk (eds), Narodowy Korpus Języka Polskiego, pp. 51–58. Warszawa: Wydawnictwo Naukowe PWN. ISBN 978-83-01-16700-4.

– Adrian Zasina, Michal Škrabal