AplikaceAplikace
Nastavení

Differences

This shows you the differences between two versions of the page.

Link to this comparison view

Next revision
Previous revision
Last revisionBoth sides next revision
en:cnk:nkjp [2018/11/02 13:05] – created Adrian Zasinaen:cnk:nkjp [2018/11/06 10:33] – [Corpus NKJP_1M] numbers Adrian Zasina
Line 2: Line 2:
 ====== Corpus NKJP_1M ====== ====== Corpus NKJP_1M ======
  
-NKJP_1M corpus is a manually annotated one-milion word subcorpus of the [[http://nkjp.pl| National Corpus of Polish]] (NKJP – //Narodowy Korpus Języka Polskiego//compiled 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. Corpus has lemmatisation and morphological annotation; bearing in mind text categorisation is representative. +The NKJP_1M corpus is a manually annotated one million word subcorpus of the [[http://nkjp.pl| 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 lemmatisationmorphological annotation, and representative coverage of text categories.
  
 <WRAP right 35%> <WRAP right 35%>
 ^ <fs medium>Name</fs> ^^ <fs medium>NKJP_1M</fs> ^ ^ <fs medium>Name</fs> ^^ <fs medium>NKJP_1M</fs> ^
-^ Positions ^ Number of positions (tokens) |  1 215 513 |   +^ Positions ^ Number of positions (tokens) |  1,215,513 |   
-^ ::: ^ Number of positions (excl. punctuation) |  992 014 |   +^ ::: ^ Number of positions (excl. punctuation) |  992,014 |   
-^ ::: ^ Number of word forms |  143 477 |   +^ ::: ^ Number of word forms |  143,477 |   
-^ ::: ^ Number of lemmas |  54 174 | +^ ::: ^ Number of lemmas |  54,174 | 
-^ Structures ^ Number of documents <doc> |  3 889 | +^ Structures ^ Number of documents <doc> |  3,889 | 
-^ ::: ^ Number of paragraphs <p> |  18 484 | +^ ::: ^ Number of paragraphs <p> |  18,484 | 
-^ ::: ^ Number of sentences <s> |  85 663 | +^ ::: ^ Number of sentences <s> |  85,663 | 
-^ Further information ^ Reference corpus |  ANO |   +^ Further information ^ Reference corpus |  YES |   
-^ ::: ^ Representative corpus |  ANO |+^ ::: ^ Representative corpus |  YES |
 ^ ::: ^ Publication year |  2018 | ^ ::: ^ Publication year |  2018 |
 </WRAP> </WRAP>
  
 ===== Text classification ===== ===== Text classification =====
-Text classification of NKJP_1M combines traditional and thematic-genre text categorisationText categorisation into genre (in Polish corpus terminology rather //type//) is often integrated with the medium (in Polish corpus terminology rather //communication channel//) categorisationnevertheless these two type of categorisation are separate. + 
 +The classification of NKJP_1M texts combines traditional criteria with a categorisation based on topic and genre. Genre categories (or //types//, in Polish corpus terminologyare 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 ^ ^Communication layer^ doc.genre ^ Category ^ Proportion ^
-| written | #typ_publ | journalism |  48,85 %| +| written | #typ_publ | journalism |  48.85%| 
-| ::: | #typ_lit | fiction |  17,04 %| +| ::: | #typ_lit | fiction |  17.04%| 
-| ::: | #typ_fakt | non-fiction |  5,34 %| +| ::: | #typ_fakt | non-fiction |  5.34%| 
-| ::: | #typ_inf-por | informative type |  5,62 %| +| ::: | #typ_inf-por | informative texts |  5.62%| 
-| ::: | #typ_urzed | legal texts |  2,97 %| +| ::: | #typ_urzed | legal texts |  2.97%| 
-| ::: | #typ_nd | scientific and teaching texts |  1,91 %| +| ::: | #typ_nd | popular science texts |  1.91%| 
-| ::: | #typ_nklas | non-fiction unclassified book |  1,00 %| +| ::: | #typ_nklas | non-fiction unclassified book |  1.00%| 
-| ::: | #typ_listy | correspondence|  0,04 %| +| ::: | #typ_listy | correspondence|  0.04%| 
-| ::: | #typ_lit_poezja | poetry |  0,01 %| +| ::: | #typ_lit_poezja | poetry |  0.01%| 
-| spoken | #typ_qmow | quasi-spoken texts |  2,50 %| +| spoken | #typ_qmow | quasi-spoken texts |  2.50%| 
-| ::: | #typ_media | spoken media text |  2,07 %| +| ::: | #typ_media | spoken media text |  2.07%| 
-| ::: | #typ_konwers | spoken conversational texts |  5,57 %| +| ::: | #typ_konwers | spoken conversational texts |  5.57%| 
-| web | #typ_net_interakt | dynamic Internet texts |  5,18 %| +| web | #typ_net_interakt | interaction-based Internet texts |  5.18%| 
-| ::: | #typ_net_nieinterakt | static Internet texts |  1,91 %|+| ::: | #typ_net_nieinterakt | non-interaction-based Internet texts |  1.91%|
  
 ===== Positional annotation and tagging ===== ===== Positional annotation and tagging =====
  
-Compared to the Czech corpora NKJP_1M has in addition a positional attribute which is specific for Polish, so called **flexeme**.  It is a category based on part of speeches that are further divided into more specific lexeme classes. Thusthere are for example within nouns (//subst//differed depreciative nouns (//depr//), beside of common adjectives (//adj//there are ad-adjectival adjectives (//adja//, e.g. //__biało__-czerwony//, //__sportowo__-rekreacyjny//), post-prepositional adjectives (//adjp//, e.g. //po __polsku__//, //od __dawna__//), predicative adjectives (//adjc//, e.g. //jestem __pewien__//, //był __wesół__ i __zdrów__//), particulary is delicate the distinction of verbal categories (more than 10 different flexemes). +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 list of all flexemes) is available [[http://nkjp.pl/poliqarp/help/ense2.html|here]].+Moreover, the Polish tagset differs from the Czech one; its detailed description (including the full flexeme list) is available [[http://nkjp.pl/poliqarp/help/ense2.html|here]].
  
-In addtion, two positional attributes were added to the original corpus: ''lc'' and ''lemma_lc'', which allow to search regardless of case-sensitive in the corpus. +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 ====== ====== How to cite NKJP_1M ======