Dataset Viewer
Auto-converted to Parquet Duplicate
id
stringlengths
13
13
tokens
listlengths
1
1.02k
ner_tags
listlengths
1
1.02k
train_0000001
[ "aw1805–1876", ")", ",", "who", "in", "turn", "gave", "birth", "to", "several", "children", ",", "among", "them—from", "her", "liaison", "to", "Franz", "Liszt", "–-", "Cosima", "Wagner", "(", "1837–1930", ")", "." ]
[ "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-PER", "I-PER", "O", "B-PER", "I-PER", "O", "O", "O", "O" ]
train_0000002
[ "St.", "Mary", "'s", "Catholic", "Church", "(", "Sandusky", ",", "Ohio", ")" ]
[ "B-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG" ]
train_0000003
[ "Shhh", "...", "Do", "n't", "Tell", "''" ]
[ "B-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "O" ]
train_0000004
[ "Antonín", "Rýgr", "(", "1970–72", ")" ]
[ "B-PER", "I-PER", "O", "O", "O" ]
train_0000005
[ "*Inducted", "into", "the", "United", "States", "Hockey", "Hall", "of", "Fame", "in", "2015" ]
[ "O", "O", "O", "B-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "O", "O" ]
train_0000006
[ "Dia", "Evtimova", "Catalina", "Castaño", "Dinah", "Pfizenmaier", "Tatjana", "Malek" ]
[ "B-PER", "I-PER", "B-PER", "I-PER", "B-PER", "I-PER", "B-PER", "I-PER" ]
train_0000007
[ "102.7", "MHz", "Triple", "J", "–", "ABC" ]
[ "O", "O", "B-ORG", "I-ORG", "O", "B-ORG" ]
train_0000008
[ "*20px", "Idaho", "State", "Highway", "5" ]
[ "O", "B-ORG", "I-ORG", "I-ORG", "I-ORG" ]
train_0000009
[ "''", "Atelocentra", "''", "Meyrick", ",", "1884" ]
[ "O", "B-LOC", "O", "O", "O", "O" ]
train_0000010
[ "'", "''", "Antarctic", "prion", "''", "'" ]
[ "O", "O", "B-LOC", "I-LOC", "O", "O" ]
train_0000011
[ "James", "Franco", "(", "M.F.A", "." ]
[ "B-PER", "I-PER", "O", "O", "O" ]
train_0000012
[ "Prince", "Albert", "Victor", ",", "Duke", "of", "Clarence", "and", "Avondale" ]
[ "B-PER", "I-PER", "I-PER", "I-PER", "I-PER", "I-PER", "I-PER", "I-PER", "I-PER" ]
train_0000013
[ "How", "to", "Get", "Away", "with", "Murder", "''" ]
[ "B-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "O" ]
train_0000014
[ "Durrães", "e", "Tregosa" ]
[ "B-LOC", "I-LOC", "I-LOC" ]
train_0000015
[ "Brent", "Cross", "tube", "station" ]
[ "B-ORG", "I-ORG", "I-ORG", "I-ORG" ]
train_0000016
[ "First", "recorded", "in", "the", "Serranía", "de", "las", "Quinchas", "on", "January", "17", ",", "2006", "." ]
[ "O", "O", "O", "O", "B-LOC", "I-LOC", "I-LOC", "I-LOC", "O", "O", "O", "O", "O", "O" ]
train_0000017
[ "Civic", "Freedom", "Party" ]
[ "B-ORG", "I-ORG", "I-ORG" ]
train_0000018
[ "キイロキリガ", "—", "Xanthia", "togata", "''", "(", "Esper", ",", "1788", ")" ]
[ "O", "O", "B-LOC", "I-LOC", "O", "O", "O", "O", "O", "O" ]
train_0000019
[ "*August", "8", "–", "Craig", "Breslow" ]
[ "O", "O", "O", "B-PER", "I-PER" ]
train_0000020
[ "He", "has", "also", "won", "the", "Japan", "Senior", "Open", "twice", "." ]
[ "O", "O", "O", "O", "O", "B-ORG", "I-ORG", "I-ORG", "O", "O" ]
train_0000021
[ "His", "great-grandson", "is", "director", "Whit", "Stillman", "." ]
[ "O", "O", "O", "O", "B-PER", "I-PER", "O" ]
train_0000022
[ "''", "'", "''", "Chauliodus", "''", "'", "''" ]
[ "O", "O", "O", "B-ORG", "O", "O", "O" ]
train_0000023
[ "Janaki]]", ",", "S.", "P.", "Balasubrahmanyam" ]
[ "O", "O", "B-PER", "I-PER", "I-PER" ]
train_0000024
[ ":", "'", "''", "Thornhill", "Islanders", "''", "'", "defeated", "Markham", "Waxers", "''4-games-to-3", "''" ]
[ "O", "O", "O", "B-ORG", "I-ORG", "O", "O", "O", "B-ORG", "I-ORG", "O", "O" ]
train_0000025
[ "Obituary", "of", "Juliette", "Peirce", "." ]
[ "O", "O", "B-PER", "I-PER", "O" ]
train_0000026
[ "78", "''", "in", "Osnabrück", "." ]
[ "O", "O", "O", "B-LOC", "O" ]
train_0000027
[ "Gian", "Girolamo", "Albani" ]
[ "B-PER", "I-PER", "I-PER" ]
train_0000028
[ "``", "Rhinestone", "Cowboy", "''", "(", "Larry", "Weiss", ")", "-", "3:15" ]
[ "O", "B-ORG", "I-ORG", "O", "O", "B-PER", "I-PER", "O", "O", "O" ]
train_0000029
[ "*Episode", "9", "-", "Michael", "Conlan" ]
[ "O", "O", "O", "B-PER", "I-PER" ]
train_0000030
[ "Antawn", "Jamison", "(", "28", ")" ]
[ "B-PER", "I-PER", "O", "O", "O" ]
train_0000031
[ "Chillicothe", "Paints", "(", "1993–2008", ",", "joined", "the", "Prospect", "League", ")" ]
[ "B-ORG", "I-ORG", "O", "O", "O", "O", "O", "B-ORG", "I-ORG", "O" ]
train_0000032
[ "Antoine", "Dadin", "de", "Hauteserre" ]
[ "B-PER", "I-PER", "I-PER", "I-PER" ]
train_0000033
[ ":", "'", "''", "Fifth", "Brother", "(", "portrayed", "by", "Philip", "Anthony-Rodriguez", ")", "'", "''" ]
[ "O", "O", "O", "O", "O", "O", "O", "O", "B-PER", "I-PER", "O", "O", "O" ]
train_0000034
[ "Château", "de", "Champs-sur-Marne" ]
[ "B-ORG", "I-ORG", "I-ORG" ]
train_0000035
[ "Secretary-General", "of", "the", "United", "Nations" ]
[ "B-LOC", "I-LOC", "I-LOC", "I-LOC", "I-LOC" ]
train_0000036
[ "1978", "demo", "released", "on", "Off", "the", "Wall", "(", "Special", "Edition", ")", "''", "(", "2001", ")" ]
[ "O", "O", "O", "O", "B-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "O", "O", "O", "O" ]
train_0000037
[ "North", "Fanling", "New", "Town", "(", "Extension", "of", "Fanling-Sheung", "Shui", "New", "Town", ")" ]
[ "B-LOC", "I-LOC", "I-LOC", "I-LOC", "O", "O", "O", "B-LOC", "I-LOC", "I-LOC", "I-LOC", "O" ]
train_0000038
[ "Carina", "Witthöft", ",", "6–3", ",", "7–5" ]
[ "B-PER", "I-PER", "O", "O", "O", "O" ]
train_0000039
[ "File", ":", "CCStaClara.JPG|Through", "Santa", "Clara" ]
[ "O", "O", "O", "B-LOC", "I-LOC" ]
train_0000040
[ "Eveno-Bytantaysky", "National", "District" ]
[ "B-LOC", "I-LOC", "I-LOC" ]
train_0000041
[ "''", "Euxoa", "centralis", "''", "(", "Staudinger", ",", "1889", ")" ]
[ "O", "B-ORG", "I-ORG", "O", "O", "O", "O", "O", "O" ]
train_0000042
[ "Diocese", "of", "San", "Andrés", "Tuxtla" ]
[ "B-LOC", "I-LOC", "I-LOC", "I-LOC", "I-LOC" ]
train_0000043
[ "Palisades", ",", "New", "York" ]
[ "B-LOC", "I-LOC", "I-LOC", "I-LOC" ]
train_0000044
[ "Rosephanye", "Powell", "(", "born", "1962", ")" ]
[ "B-PER", "I-PER", "O", "O", "O", "O" ]
train_0000045
[ "Chum", "Ta", "Bong", "District" ]
[ "B-LOC", "I-LOC", "I-LOC", "I-LOC" ]
train_0000046
[ "Epica", "–", "Consign", "to", "Oblivion", "''" ]
[ "B-ORG", "O", "B-ORG", "I-ORG", "I-ORG", "O" ]
train_0000047
[ "List", "of", "Flywheel", ",", "Shyster", ",", "and", "Flywheel", "(", "1990", "radio", "series", ")", "episodes" ]
[ "B-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG" ]
train_0000048
[ "''", "Epirrhoe", "rivata", "''", "(", "Hübner", ",", "1813", ")" ]
[ "O", "B-LOC", "I-LOC", "O", "O", "O", "O", "O", "O" ]
train_0000049
[ "**", "''", "Justice", "League", "Elite", "''", ":" ]
[ "O", "O", "B-ORG", "I-ORG", "I-ORG", "O", "O" ]
train_0000050
[ "It", "is", "found", "in", "Peru", "." ]
[ "O", "O", "O", "O", "B-LOC", "O" ]
train_0000051
[ "**", "Ecuador", ":", "7", "%", "of", "the", "population" ]
[ "O", "B-LOC", "O", "O", "O", "O", "O", "O" ]
train_0000052
[ "Translator", "of", "KCCV-FM", ",", "Overland", "Park", ",", "Kansas" ]
[ "O", "O", "B-ORG", "O", "B-LOC", "I-LOC", "I-LOC", "I-LOC" ]
train_0000053
[ "Mrs", "Brigitte", "Vallée", "." ]
[ "O", "B-PER", "I-PER", "O" ]
train_0000054
[ "Lipiny", ",", "Ciechanów", "County" ]
[ "B-LOC", "I-LOC", "I-LOC", "I-LOC" ]
train_0000055
[ "Tatiana", "Volosozhar", "/", "Maxim", "Trankov" ]
[ "B-PER", "I-PER", "O", "B-PER", "I-PER" ]
train_0000056
[ "Raymerville", "–", "Markville", "East", ",", "Ontario" ]
[ "B-LOC", "I-LOC", "I-LOC", "I-LOC", "I-LOC", "I-LOC" ]
train_0000057
[ "Robert", "Maxwell", ",", "5th", "Lord", "Maxwell", "(", "1493–1546", ")" ]
[ "B-PER", "I-PER", "I-PER", "I-PER", "I-PER", "I-PER", "O", "O", "O" ]
train_0000058
[ "Prince", "Rupert", "of", "the", "Rhine", "15", "August", "1672", "–", "19", "November", "1682" ]
[ "B-PER", "I-PER", "I-PER", "I-PER", "I-PER", "O", "O", "O", "O", "O", "O", "O" ]
train_0000059
[ "Železničná", "spoločnosť", "Slovensko" ]
[ "B-ORG", "I-ORG", "I-ORG" ]
train_0000060
[ "On", "February", "6", ",", "1925", "he", "retired", "from", "active", "duty", "and", "died", "in", "Santiago", "in", "1938", "." ]
[ "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-LOC", "O", "O", "O" ]
train_0000061
[ "The", "67", "'s", "finished", "in", "last", "place", "in", "the", "2012–13", "OHL", "season", "with", "just", "16", "wins", "." ]
[ "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-ORG", "I-ORG", "I-ORG", "O", "O", "O", "O", "O" ]
train_0000062
[ "Internet", "Exchange", "Point", "(", "IXP", ")" ]
[ "B-ORG", "I-ORG", "I-ORG", "O", "O", "O" ]
train_0000063
[ "``", "Suzukake", "no", "Ki", "no", "Michi", "de", "``", "Kimi", "no", "Hohoemi", "o", "Yume", "ni", "Miru", "''", "to", "Itte", "Shimattara", "Bokutachi", "no", "Kankei", "wa", "Dō", "Kawatte", "Shimau", "no", "ka", ",", "Bokunari", "ni", "Nan-nichi", ...
[ "O", "B-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "...
train_0000064
[ "It", "was", "released", "as", "the", "first", "single", "from", "their", "2004", "album", "Everybody", "Loves", "a", "Happy", "Ending", "''", "and", "is", "the", "first", "single", "that", "features", "both", "original", "members", "Roland", "Orzabal", "and...
[ "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-PER", "I-PER", "O", "B-PER", "I-PER", "O", "O", "O" ]
train_0000065
[ "Conservative]]", "Leader", "David", "Cameron", ")" ]
[ "O", "O", "B-PER", "I-PER", "O" ]
train_0000066
[ "Much", "of", "the", "story", "takes", "place", "in", "Flagstaff", ",", "Arizona", ",", "where", "Sgt", "." ]
[ "O", "O", "O", "O", "O", "O", "O", "B-LOC", "I-LOC", "I-LOC", "O", "O", "O", "O" ]
train_0000067
[ "''", "Invisible", "Touch", "''", "–", "Genesis" ]
[ "O", "B-ORG", "I-ORG", "O", "O", "B-ORG" ]
train_0000068
[ "United", "States", "Senate", "Librarian" ]
[ "B-PER", "I-PER", "I-PER", "I-PER" ]
train_0000069
[ "Live", "at", "Sin-é", "(", "Legacy", "Edition", ")", "''" ]
[ "B-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "O" ]
train_0000070
[ "'", "''", "Calgary", "''", "'", ",", "'", "''", "Canada", "''", "'" ]
[ "O", "O", "B-LOC", "O", "O", "O", "O", "O", "B-LOC", "O", "O" ]
train_0000071
[ "*2001", "–", "Hugh", "Hefner", ",", "Roastmaster", "Jimmy", "Kimmel" ]
[ "O", "O", "B-PER", "I-PER", "O", "O", "B-PER", "I-PER" ]
train_0000072
[ "Aigle", "Noir", "AC" ]
[ "B-ORG", "I-ORG", "I-ORG" ]
train_0000073
[ "Hathaways", "Mountain", "Pines", ",", "California" ]
[ "B-LOC", "I-LOC", "I-LOC", "I-LOC", "I-LOC" ]
train_0000074
[ "Peter", "T.", "King" ]
[ "B-PER", "I-PER", "I-PER" ]
train_0000075
[ "Charlie", "Clouser", "–", "Nine", "Inch", "Nails" ]
[ "B-PER", "I-PER", "O", "B-ORG", "I-ORG", "I-ORG" ]
train_0000076
[ "Former", "collegiate", "church", "St", "Bartholomew", ",", "Liège" ]
[ "B-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "O", "B-LOC" ]
train_0000077
[ "It", "was", "created", "on", "28", "January", "1831", "for", "Standish", "O'Grady", ",", "Attorney-General", "for", "Ireland", "and", "Lord", "Chief", "Baron", "of", "the", "Exchequer", "for", "Ireland", "." ]
[ "O", "O", "O", "O", "O", "O", "O", "O", "B-PER", "I-PER", "O", "B-ORG", "I-ORG", "I-ORG", "O", "B-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "O" ]
train_0000078
[ "Zlaté", "Moravce", "District" ]
[ "B-LOC", "I-LOC", "I-LOC" ]
train_0000079
[ "Wilhelm", "Gustav", "Franz", "Herter" ]
[ "B-PER", "I-PER", "I-PER", "I-PER" ]
train_0000080
[ "Thomasville", ",", "Missouri" ]
[ "B-LOC", "I-LOC", "I-LOC" ]
train_0000081
[ "Clement", "Poitrenaud", ",", "14", "." ]
[ "B-ORG", "I-ORG", "O", "O", "O" ]
train_0000082
[ "Yahn", "|", "Yazdan", "Chah", "|", "Yegi", "|", "Yekkeh", "Derakht", "|", "Yosht", "|", "Yusefabad", "|", "Yush", "|", "Yushu" ]
[ "B-LOC", "O", "B-LOC", "I-LOC", "O", "B-LOC", "O", "B-LOC", "I-LOC", "O", "B-LOC", "O", "B-LOC", "O", "B-LOC", "O", "B-LOC" ]
train_0000083
[ "1982", "Tour", "de", "France" ]
[ "B-ORG", "I-ORG", "I-ORG", "I-ORG" ]
train_0000084
[ "John", "Denison-Pender", ",", "1st", "Baron", "Pender" ]
[ "B-PER", "I-PER", "I-PER", "I-PER", "I-PER", "I-PER" ]
train_0000085
[ "''", "Glyphipterix", "iometalla", "''", "(", "Meyrick", ",", "1880", ")" ]
[ "O", "B-LOC", "I-LOC", "O", "O", "O", "O", "O", "O" ]
train_0000086
[ "''", "Zygaena", "cynarae", "''", "(", "Esper", ",", "1789", ")" ]
[ "O", "B-LOC", "I-LOC", "O", "O", "O", "O", "O", "O" ]
train_0000087
[ "''", "Super", "World", "of", "Sports", "''", "'" ]
[ "O", "B-ORG", "I-ORG", "I-ORG", "I-ORG", "O", "O" ]
train_0000088
[ "Chananel", "ben", "Chushiel" ]
[ "B-PER", "I-PER", "I-PER" ]
train_0000089
[ "Bobby", "Rush", "(", "D", ")" ]
[ "B-PER", "I-PER", "O", "O", "O" ]
train_0000090
[ "Legend", "of", "Chun", "Hyang", "''" ]
[ "B-ORG", "I-ORG", "I-ORG", "I-ORG", "O" ]
train_0000091
[ "BBC", "(", "Test", "Match", "Special", "''", ")" ]
[ "B-ORG", "O", "B-ORG", "I-ORG", "I-ORG", "O", "O" ]
train_0000092
[ "Bruce", "Beresford", "(", "Non-Jew", ")" ]
[ "B-PER", "I-PER", "O", "O", "O" ]
train_0000093
[ "It", "was", "also", "released", "along", "with", "Visual", "Studio", "2005", ",", "Microsoft", "SQL", "Server", "2005", ",", "and", "BizTalk", "2006", "." ]
[ "O", "O", "O", "O", "O", "O", "B-ORG", "I-ORG", "I-ORG", "O", "B-ORG", "I-ORG", "I-ORG", "I-ORG", "O", "O", "B-ORG", "O", "O" ]
train_0000094
[ "Leonardo", "Rodríguez", "Solís" ]
[ "B-PER", "I-PER", "I-PER" ]
train_0000095
[ "She", "was", "born", "in", "Aldsworth", ",", "Gloucestershire", ",", "England", "on", "10", "January", "1836", "." ]
[ "O", "O", "O", "O", "B-LOC", "O", "B-LOC", "O", "B-LOC", "O", "O", "O", "O", "O" ]
train_0000096
[ "**", "Kyoto", "National", "Museum", "(", "early", "production", "cast", ")" ]
[ "O", "B-ORG", "I-ORG", "I-ORG", "O", "O", "O", "O", "O" ]
train_0000097
[ "List", "of", "years", "in", "Brazil" ]
[ "B-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG" ]
train_0000098
[ "The", "International", "League", "of", "Dermatological", "Societies" ]
[ "B-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG", "I-ORG" ]
train_0000099
[ "In", "June", "1999", "it", "was", "sold", "to", "National", "Express", "who", "in", "September", "2004", "sold", "it", "to", "Connex", "." ]
[ "O", "O", "O", "O", "O", "O", "O", "B-ORG", "I-ORG", "O", "O", "O", "O", "O", "O", "O", "B-ORG", "O" ]
train_0000100
[ "Later", "reductions", "were", "extended", "into", "the", "areas", "that", "correspond", "to", "Argentina", ",", "Brazil", ",", "Bolivia", "and", "Uruguay", "." ]
[ "O", "O", "O", "O", "O", "O", "O", "O", "O", "O", "B-LOC", "O", "B-LOC", "O", "B-LOC", "O", "B-LOC", "O" ]
End of preview. Expand in Data Studio

KRW Unified Multilingual 47-Class Named Entity Recognition Dataset

Curated & Published by: Rishabh Kumar (@Rishabh157)

KRW Unified Multilingual NER is a master, large-scale multilingual dataset containing 2,252,976 sentences (~47.4 million tokens) across 47 fine-grained entity classes in 15+ languages and diverse scripts (Latin, Cyrillic, Indic, and East Asian).

By harmonizing and uniting three foundational NER corpora—WikiANN, MultiNERD, and MultiCoNER 2023—this dataset bridges the gap between classic coarse-grained entities (PER, LOC, ORG), complex low-context web queries, and specialized biomedical, cultural, and technical entity categories.


📊 Dataset Summary & Key Statistics

Split Sentences Total Tokens Approx. Entity Mentions Format
Train 1,530,020 33,506,270 ~2,325,000 Parquet (2 Shards)
Validation 186,295 4,058,776 ~280,000 Parquet (1 Shard)
Test 536,661 9,828,034 ~760,000 Parquet (1 Shard)
Total 2,252,976 47,393,080 ~3,365,000 Apache Parquet (231.7 MB)
  • Average Sentence Length: 21.0 tokens
  • 95th Percentile Span Length: ≤ 5 tokens
  • 99th Percentile Span Length: ≤ 9 tokens
  • Total Unique Entity Classes: 47 (95 unique BIO tags: 47 B-, 47 I-, and 1 O)

🏷️ Entity Taxonomy (47 Fine-Grained Classes)

The dataset provides fine-grained entity classification spanning 7 distinct semantic domains:

1. People & Roles (8 Classes)

  • PER: General person mentions
  • OtherPER: Non-canonical, fictional, or ambiguous person entities
  • Artist: Painters, sculptors, designers, creators
  • Athlete: Sports players, Olympians, drivers
  • Cleric: Religious figures, leaders, priests
  • Politician: Presidents, ministers, senators, diplomats
  • Scientist: Researchers, physicists, mathematicians, scholars
  • SportsManager: Coaches, referees, sports team managers

2. Locations & Infrastructure (5 Classes)

  • LOC: General geographical locations
  • OtherLOC: Natural geographic features (mountains, lakes, rivers)
  • HumanSettlement: Cities, towns, villages, municipalities
  • Facility: Stadiums, airports, bridges, monuments, campuses
  • Station: Transit stations, subway stops, railway junctions

3. Organizations & Corporations (7 Classes)

  • ORG: General organizations, NGOs, institutions, bodies
  • PublicCorp: Publicly traded corporations (e.g., Apple, Microsoft)
  • PrivateCorp: Private enterprises, startups, firms
  • AerospaceManufacturer: Aircraft, space exploration, rocket manufacturers
  • CarManufacturer: Automotive corporations
  • SportsGRP: Athletic clubs, sports franchises, national teams
  • MusicalGRP: Bands, orchestras, musical ensembles

4. Biomedical, Health & Science (7 Classes)

  • AnatomicalStructure: Organs, bones, tissues, biological systems
  • Disease: Pathologies, medical conditions, syndromes
  • DIS: Broad medical disorders
  • Symptom: Clinical signs and physical symptoms
  • Medication/Vaccine: Pharmaceuticals, vaccines, active compounds
  • MedicalProcedure: Surgeries, diagnostic tests, clinical therapies
  • BIO: Biochemical compounds and molecules

5. Creative, Media & Cultural Works (6 Classes)

  • ArtWork: Sculptures, paintings, monuments
  • VisualWork: Films, TV shows, documentaries
  • WrittenWork: Books, novels, poems, research papers
  • MusicalWork: Songs, symphonies, albums
  • MEDIA: Broadcasting networks, magazines, news publications
  • MYTH: Mythological and legendary figures/deities

6. Products, Goods & Transportation (8 Classes)

  • Software: Operating systems, apps, algorithms, digital platforms
  • Vehicle: Automobiles, ships, trains, aircraft
  • VEHI: General mechanical vehicles
  • Clothing: Garments, apparel brands, fashion items
  • Drink: Beverages, cocktails, wines, soft drinks
  • Food: Dishes, recipes, culinary items
  • FOOD: Food products and ingredients
  • INST: Scientific, musical, or technical instruments
  • OtherPROD: Commercial products not covered elsewhere

7. Nature, Biology & Miscellaneous (6 Classes)

  • ANIM: Animals, biological species, breeds
  • PLANT: Flora, trees, botanical species
  • CEL: Celestial bodies (stars, planets, galaxies, asteroids)
  • EVE: Historical, sporting, or political events
  • TIME: Explicit temporal expressions

🌐 Language Coverage & Writing Systems

The dataset encompasses 15+ distinct languages spanning 5 major language families and 6 different scripts, providing true cross-lingual syntactic and typographical diversity:

Language ISO Code Writing Script Language Family Source Corpora Typographical & Linguistic Features
English en Latin Indo-European (Germanic) WikiANN, MultiNERD, MultiCoNER Baseline global entity distributions & encyclopedic text
German de Latin (with Umlauts) Indo-European (Germanic) MultiNERD, MultiCoNER Standard noun capitalization; multi-word compound nouns
Dutch nl Latin Indo-European (Germanic) MultiNERD Germanic syntax with complex prepositional titles
Swedish sv Latin Indo-European (North Germanic) MultiCoNER Scandinavian orthography; compound noun phrases
Spanish es Latin Indo-European (Romance) MultiNERD, MultiCoNER Post-nominal adjectival modifiers; accent markers
French fr Latin Indo-European (Romance) MultiNERD, MultiCoNER Elision articles (l', d'), accents, and hyphens
Italian it Latin Indo-European (Romance) MultiNERD, MultiCoNER Articulated prepositions (dell', negli)
Portuguese pt Latin Indo-European (Romance) MultiNERD, MultiCoNER Nasal vowels, tildes, and compound surnames
Russian ru Cyrillic Indo-European (Slavic) MultiNERD Rich inflectional morphology and multi-case inflections
Polish pl Latin (Extended) Indo-European (Slavic) MultiNERD Complex consonant clusters and case endings
Ukrainian uk Cyrillic Indo-European (Slavic) MultiCoNER East Slavic orthography, apostrophes, and declensions
Hindi hi Devanagari Indo-European (Indo-Aryan) MultiCoNER Matra vowel diacritics, conjuncts, zero-capitalization
Bengali bn Bengali-Assamese Indo-European (Indo-Aryan) MultiCoNER Eastern Indo-Aryan script; agglutinative case suffixes
Farsi / Persian fa Perso-Arabic (RTL) Indo-European (Iranian) MultiCoNER Right-to-Left script, zero-width non-joiners (ZWNJ)
Chinese (Simplified) zh Hanzi (Logographic) Sino-Tibetan (Sinitic) MultiNERD, MultiCoNER Continuous unspaced text; character-level span boundaries
Multilingual (MULTI) MULTI Mixed / Code-Switched Multi-family MultiCoNER Ambiguous web queries, multilingual search snippets

Key Cross-Lingual Capabilities

  1. Script Diversity: Models trained on this dataset learn cross-lingual transfer across Latin, Cyrillic, Devanagari, Bengali, Perso-Arabic, and Hanzi glyphs.
  2. Casing Invariance: Incorporates both case-sensitive languages (English, German with capitalized nouns) and casing-free scripts (Hindi, Bengali, Chinese, Farsi) where models cannot rely on initial capital letters to detect named entities.
  3. Segmentation Robustness: Combines whitespace-delimited languages with character-delimited languages (Chinese) and morphologically rich inflected languages (Russian, Polish).

📦 Data Sources, Provenance & Composition

The dataset unifies three foundational NER benchmarks, each downloaded and extracted from its official upstream distribution:

Source Benchmark Upstream Origin / Repository Subsets Downloaded & Merged Primary Focus Train Sentences Val Sentences Test Sentences
MultiNERD Babelscape/multinerd 10 languages: en, de, es, fr, it, nl, pl, pt, ru, zh Multi-genre encyclopedic & news text across 15 fine-grained entity types 1,339,200 162,697 236,514
MultiCoNER 2023
(SemEval-2023 Task 2)
Amazon AWS S3
https://multiconer.s3.amazonaws.com/multiconer2023
MULTI-Multilingual code-mixed track Short, low-context, complex search queries and acoustic-noisy web text 170,824 13,598 290,147
WikiANN (PAN-X) wikiann English (en) standard benchmark split Encyclopedic Wikipedia anchor for canonical PER, LOC, ORG entities 20,000 10,000 10,000
KRW Unified Total Harmonized 47-Class Master Corpus 15+ Languages across 6 Scripts Unified Multilingual NER 1,530,020 186,295 536,661

Data Preparation & Harmonization Pipeline

  1. Download & Extraction: MultiNERD and WikiANN were fetched via Hugging Face datasets, while MultiCoNER 2023 was retrieved from its official AWS S3 distribution.
  2. Text Normalization: Unicode NFKC normalization applied to eliminate invisible control characters, irregular whitespace, and malformed glyphs.
  3. BIO Sequence Auto-Repair: Heuristic repair applied to eliminate illegal BIO sequence breaks (e.g., orphan I- tags without preceding B- tags converted to B-).
  4. Taxonomy Harmonization: All entity tags mapped into a unified 95-tag BIO schema (47 classes $\times$ 2 (B-, I-) + O).
  5. Deterministic Serialization: Exported into compressed Apache Parquet shards with 64-bit integer sentence counts and unique global IDs (train_XXXXXXX, val_XXXXXXX, test_XXXXXXX).

🚀 Quickstart Usage

1. Load with Hugging Face datasets

from datasets import load_dataset

# Load the full unified dataset
dataset = load_dataset("Rishabh157/krw-unified-multilingual-ner")

# Inspect dataset splits
print(dataset)
# Output:
# DatasetDict({
#     train: Dataset({features: ['id', 'tokens', 'ner_tags'], num_rows: 1530020}),
#     validation: Dataset({features: ['id', 'tokens', 'ner_tags'], num_rows: 186295}),
#     test: Dataset({features: ['id', 'tokens', 'ner_tags'], num_rows: 536661})
# })

# View a single example
sample = dataset["train"][0]
print("Tokens:", sample["tokens"])
print("Tags:  ", sample["ner_tags"])

2. Streaming Mode (Zero-RAM Overhead)

For quick exploration or environments with limited memory:

dataset_stream = load_dataset(
    "Rishabh157/krw-unified-multilingual-ner",
    split="train",
    streaming=True
)

for example in dataset_stream.take(5):
    print(" ".join(example["tokens"]))

3. Standard Token Classification (Hugging Face Transformers)

The dataset is framework-agnostic and works directly with any standard Hugging Face pipeline, Transformers model (AutoModelForTokenClassification), SpaCy, GLiNER, or custom PyTorch loop:

from transformers import AutoTokenizer, AutoModelForTokenClassification
from datasets import load_dataset

# 1. Load dataset
dataset = load_dataset("Rishabh157/krw-unified-multilingual-ner")

# 2. Tokenize using any pretrained multilingual backbone
tokenizer = AutoTokenizer.from_pretrained("microsoft/mdeberta-v3-base")

def tokenize_and_align_labels(examples):
    tokenized_inputs = tokenizer(
        examples["tokens"],
        truncation=True,
        is_split_into_words=True,
        max_length=256
    )
    return tokenized_inputs

encoded_dataset = dataset.map(tokenize_and_align_labels, batched=True)

💾 Data Storage Architecture, Formats & Sharding

To optimize for large-scale GPU training throughput, low bandwidth transfers, interactive browser inspection, and broad library interoperability, this dataset uses a hybrid storage architecture:

Repository File Layout on Hugging Face Hub:
Rishabh157/krw-unified-multilingual-ner/
├── README.md                          <-- Comprehensive Dataset Card & Metadata
└── data/                              <-- Compressed Columnar Apache Parquet Shards (231.7 MB total)
    ├── train-00000-of-00002.parquet   <-- Train Shard 1 (800,000 sentences, 84.67 MB)
    ├── train-00001-of-00002.parquet   <-- Train Shard 2 (730,020 sentences, 69.36 MB)
    ├── validation-00000-of-00001.parquet <-- Validation Split (186,295 sentences, 19.33 MB)
    └── test-00000-of-00001.parquet       <-- Test Split (536,661 sentences, 58.33 MB)

1. Primary Hub Storage: Apache Parquet

All splits on the Hugging Face Hub are serialized as Apache Parquet files compressed using Snappy:

  • Columnar Binary Layout: Instead of sequentially scanning text strings line-by-line, tokens and tags are stored in contiguous binary column buffers.
  • Massive File Compression: Reduces raw text volume from ~380 MB down to 231.7 MB in highly optimized, indexed columnar Parquet format.
  • Memory-Mapped Streaming (mmap): Hugging Face datasets streams Parquet batches directly into GPU memory without loading the entire 47M-token corpus into system RAM.
  • Native Web Dataset Viewer: Hugging Face's backend automatically parses these Parquet files, providing an instant interactive table viewer for browsing and filtering directly on the dataset page.
  • Balanced Sharding: The 1.53M training sentences are divided into two equal shards (~800k rows each) to prevent network timeout during downloads and enable multi-threaded data loading across worker processes.

📜 Dataset Structure, Schema & Type Definitions

Column Specifications

Column Name Arrow Type Python Type Description
id pa.string() str Globally unique, deterministic sentence identifier with split prefix (e.g., "train_0000001", "val_0000042").
tokens pa.list_(pa.string()) list[str] Ordered list of pre-tokenized word tokens, punctuation marks, and symbols forming the input sentence.
ner_tags pa.list_(pa.string()) list[str] List of identical length containing the corresponding BIO entity tags from the 95-tag taxonomy.

Concrete Real-World Data Record Examples

Example 1: Multi-Entity Biographical Sentence (English)

{
  "id": "train_0000001",
  "tokens": [
    "who", "in", "turn", "gave", "birth", "to", "several", "children", ",",
    "among", "them—from", "her", "liaison", "to", "Franz", "Liszt", "–-", 
    "Cosima", "Wagner", "(", "1837–1930", ")", "."
  ],
  "ner_tags": [
    "O", "O", "O", "O", "O", "O", "O", "O", "O",
    "O", "O", "O", "O", "O", "B-PER", "I-PER", "O", 
    "B-PER", "I-PER", "O", "O", "O", "O"
  ]
}

Example 2: Historical Event Mention (Spanish)

{
  "id": "train_0025110",
  "tokens": [
    "Durante", "la", "Guerra", "Civil", "la", "inmensa", "mayoría", 
    "de", "las", "imágenes", "fueron", "destruidas", "."
  ],
  "ner_tags": [
    "O", "O", "B-EVE", "I-EVE", "O", "O", "O", 
    "O", "O", "O", "O", "O", "O"
  ]
}

Example 3: Fine-Grained Clinical Entity (Persian / Farsi)

{
  "id": "val_0012401",
  "tokens": [
    "قیمت", "آنها", "نیز", "معمولاً", "از", "لوودوپا", "بیشتر", "است", "."
  ],
  "ner_tags": [
    "O", "O", "O", "O", "O", "B-Medication/Vaccine", "O", "O", "O"
  ]
}

👤 Author & Dataset Curation

This master unified dataset was compiled, harmonized, standardized, and published by Rishabh Kumar (@Rishabh157).

If you use this dataset in your research, pipelines, or models, please cite:

@misc{rishabhkumar2026_krw_unified_multilingual_ner,
  author = {Rishabh Kumar},
  title = {KRW Unified Multilingual 47-Class Named Entity Recognition Dataset},
  year = {2026},
  publisher = {Hugging Face},
  howpublished = {\url{https://huggingface.co/datasets/Rishabh157/krw-unified-multilingual-ner}}
}

⚖️ License & Foundational Attribution

This unified dataset is released under the Apache 2.0 License.

Please also acknowledge the original foundational research benchmarks that contributed to this corpus:

@inproceedings{pan-etal-2017-cross,
    title = "Cross-lingual Name Tagging and Linking for 282 Languages",
    author = "Pan, Xiaoman and Zhang, Boliang and May, Jonathan and Nothman, Joel and Knight, Kevin and Ji, Heng",
    booktitle = "Proceedings of the 55th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)",
    month = jul,
    year = "2017",
    address = "Vancouver, Canada",
    publisher = "Association for Computational Linguistics",
    pages = "1946--1958",
    url = "https://aclanthology.org/P17-1178",
    doi = "10.18653/v1/P17-1178"
}

@inproceedings{tedeschi-navigli-2022-multinerd,
    title = "{M}ulti{NERD}: A Multilingual, Multi-Genre and Fine-Grained Dataset for Named Entity Recognition (and Disambiguation)",
    author = "Tedeschi, Simone and Navigli, Roberto",
    booktitle = "Findings of the Association for Computational Linguistics: NAACL 2022",
    month = jul,
    year = "2022",
    address = "Seattle, United States",
    publisher = "Association for Computational Linguistics",
    pages = "801--812",
    url = "https://aclanthology.org/2022.findings-naacl.60",
    doi = "10.18653/v1/2022.findings-naacl.60"
}

@inproceedings{malmasi-etal-2022-semeval,
    title = "{S}em{E}val-2022 Task 11: Multilingual Complex Named Entity Recognition ({M}ulti{C}o{NER})",
    author = "Malmasi, Shervin and Fang, Anjie and Fetahu, Besnik and Kar, Sudipta and Rokhlenko, Oleg",
    booktitle = "Proceedings of the 16th International Workshop on Semantic Evaluation (SemEval-2022)",
    month = jul,
    year = "2022",
    address = "Seattle, United States",
    publisher = "Association for Computational Linguistics",
    pages = "1412--1437",
    url = "https://aclanthology.org/2022.semeval-1.196",
    doi = "10.18653/v1/2022.semeval-1.196"
}

@inproceedings{fetahu-etal-2023-semeval,
    title = "{S}em{E}val-2023 Task 2: Fine-grained Multilingual Named Entity Recognition ({M}ulti{C}o{NER} 2)",
    author = "Fetahu, Besnik and Kar, Sudipta and Chen, Zhiyu and Rokhlenko, Oleg",
    booktitle = "Proceedings of the 17th International Workshop on Semantic Evaluation (SemEval-2023)",
    month = jul,
    year = "2023",
    address = "Toronto, Canada",
    publisher = "Association for Computational Linguistics",
    pages = "2247--2265",
    url = "https://aclanthology.org/2023.semeval-1.310",
    doi = "10.18653/v1/2023.semeval-1.310"
}
Downloads last month
22

Models trained or fine-tuned on Rishabh157/krw-unified-multilingual-ner