ko_core_news_lg Open-source Korean Processing Model - Freely Deploy to Complete NLP Tasks such as Word Segmentation and Part-of-Speech Tagging

Ko Core News Lg

Developed by spacy

CPU-optimized Korean processing pipeline with complete NLP capabilities including tokenization, POS tagging, dependency parsing, named entity recognition, etc.

Sequence Labeling Korean#Korean NLP #High-precision POS tagging #Dependency parsing

Downloads 52

Release Time : 5/2/2022

Model Overview

spaCy's large Korean processing model, trained on UD Korean Kaist and KLUE datasets, supporting full NLP processing pipeline including tokenization, POS tagging, dependency parsing, named entity recognition, and more.

Model Features

Complete Korean processing pipeline

Provides full NLP processing capabilities from tokenization to named entity recognition

CPU optimization

Specially optimized for CPU environments, suitable for resource-constrained scenarios

High-quality word vectors

Includes floret word vectors (200000, 300) for better semantic understanding

Multi-task processing

Simultaneously supports various NLP tasks including POS tagging, dependency parsing, and named entity recognition

Model Capabilities

Tokenization

POS tagging

Dependency parsing

Named entity recognition

Lemmatization

Sentence boundary detection

Use Cases

Text processing

Korean text analysis

Performs grammatical analysis and structural parsing of Korean news and social media content

Accurately identifies sentence components and grammatical relationships

Information extraction

Extracts named entities (person names, locations, organization names, etc.) from Korean text

Achieves F1 score of 85.19%

Linguistic research

Korean grammar research

Analyzes Korean syntactic structures and morphological features

Provides detailed POS tagging and dependency relationship analysis

🚀 ko_core_news_lg

The Korean language processing pipeline ko_core_news_lg is optimized for CPU, offering high - performance token classification capabilities.

🚀 Quick Start

For detailed information about this model, please visit: Details

✨ Features

Optimized for CPU: Specifically designed to run efficiently on CPUs.
Multiple Components: Includes tok2vec, tagger, morphologizer, parser, lemmatizer (trainable_lemmatizer), senter, and ner.

📚 Documentation

Model Information

Property	Details
Model Type	`ko_core_news_lg`
Version	`3.7.0`
spaCy	`>=3.7.0,<3.8.0`
Default Pipeline	`tok2vec`, `tagger`, `morphologizer`, `parser`, `lemmatizer`, `attribute_ruler`, `ner`
Components	`tok2vec`, `tagger`, `morphologizer`, `parser`, `lemmatizer`, `senter`, `attribute_ruler`, `ner`
Vectors	floret (200000, 300)
Sources	UD Korean Kaist v2.8 (Choi, Jinho; Han, Na - Rae; Hwang, Jena; Chun, Jayeol) KLUE v1.1.0 (Sungjoon Park, Jihyung Moon, Sungdong Kim, Won Ik Cho, Jiyoon Han, Jangwon Park, Chisung Song, Junseong Kim, Youngsook Song, Taehwan Oh, Joohong Lee, Juhyun Oh, Sungwon Ryu, Younghoon Jeong, Inkwon Lee, Sangwoo Seo, Dongjun Lee, Hyunwoo Kim, Myeonghwa Lee, Seongbo Jang, Seungwon Do, Sunkyoung Kim, Kyungtae Lim, Jongwon Lee, Kyumin Park, Jamin Shin, Seonghyun Kim, Lucy Park, Alice Oh, Jung - Woo Ha, Kyunghyun Cho) Explosion Vectors (OSCAR 2109 + Wikipedia + OpenSubtitles + WMT News Crawl) (Explosion)
License	`CC BY - SA 4.0`
Author	Explosion

Label Scheme

View label scheme (2028 labels for 4 components)

Component	Labels
`tagger`	`_SP`, `ecs`, `etm`, `f`, `f+f+jcj`, `f+f+jcs`, `f+f+jct`, `f+f+jxt`, `f+jca`, `f+jca+jp+ecc`, `f+jca+jp+ep+ef`, `f+jca+jxc`, `f+jca+jxc+jcm`, `f+jca+jxt`, `f+jcj`, `f+jcm`, `f+jco`, `f+jcs`, `f+jct`, `f+jct+jcm`, `f+jp+ef`, `f+jp+ep+ef`, `f+jp+etm`, `f+jxc`, `f+jxt`, `f+ncn`, `f+ncn+jcm`, `f+ncn+jcs`, `f+ncn+jp+ecc`, `f+ncn+jxt`, `f+ncpa+jcm`, `f+npp+jcs`, `f+nq`, `f+xsn`, `f+xsn+jco`, `f+xsn+jxt`, `ii`, `jca`, `jca+jcm`, `jca+jxc`, `jca+jxt`, `jcc`, `jcj`, `jcm`, `jco`, `jcr`, `jcr+jxc`, `jcs`, `jct`, `jct+jcm`, `jct+jxt`, `jp+ecc`, `jp+ecs`, `jp+ef`, `jp+ef+jcr`, `jp+ef+jcr+jxc`, `jp+ep+ecs`, `jp+ep+ef`, `jp+ep+etm`, `jp+ep+etn`, `jp+etm`, `jp+etn`, `jp+etn+jco`, `jp+etn+jxc`, `jxc`, `jxc+jca`, `jxc+jco`, `jxc+jcs`, `jxt`, `mad`, `mad+jxc`, `mad+jxt`, `mag`, `mag+jca`, `mag+jcm`, `mag+jcs`, `mag+jp+ef+jcr`, `mag+jxc`, `mag+jxc+jxc`, `mag+jxt`, `mag+xsn`, `maj`, `maj+jxc`, `maj+jxt`, `mma`, `mmd`, `nbn`, `nbn+jca`, `nbn+jca+jcj`, `nbn+jca+jcm`, `nbn+jca+jp+ef`, `nbn+jca+jxc`, `nbn+jca+jxt`, `nbn+jcc`, `nbn+jcj`, `nbn+jcm`, `nbn+jco`, `nbn+jcr`, `nbn+jcs`, `nbn+jct`, `nbn+jct+jcm`, `nbn+jct+jxt`, `nbn+jp+ecc`, `nbn+jp+ecs`, `nbn+jp+ecs+jca`, `nbn+jp+ecs+jcm`, `nbn+jp+ecs+jco`, `nbn+jp+ecs+jxc`, `nbn+jp+ecs+jxt`, `nbn+jp+ecx`, `nbn+jp+ef`, `nbn+jp+ef+jca`, `nbn+jp+ef+jco`, `nbn+jp+ef+jcr`, `nbn+jp+ef+jcr+jxc`, `nbn+jp+ef+jcr+jxt`, `nbn+jp+ef+jcs`, `nbn+jp+ef+jxc`, `nbn+jp+ef+jxc+jco`, `nbn+jp+ef+jxf`, `nbn+jp+ef+jxt`, `nbn+jp+ep+ecc`, `nbn+jp+ep+ecs`, `nbn+jp+ep+ecs+jxc`, `nbn+jp+ep+ef`, `nbn+jp+ep+ef+jcr`, `nbn+jp+ep+etm`, `nbn+jp+ep+etn`, `nbn+jp+ep+etn+jco`, `nbn+jp+ep+etn+jcs`, `nbn+jp+etm`, `nbn+jp+etn`, `nbn+jp+etn+jca`, `nbn+jp+etn+jca+jxt`, `nbn+jp+etn+jco`, `nbn+jp+etn+jcs`, `nbn+jp+etn+jxc`, `nbn+jp+etn+jxt`, `nbn+jxc`, `nbn+jxc+jca`, `nbn+jxc+jca+jxc`, `nbn+jxc+jca+jxt`, `nbn+jxc+jcc`, `nbn+jxc+jcm`, `nbn+jxc+jco`, `nbn+jxc+jcs`, `nbn+jxc+jp+ef`, `nbn+jxc+jxc`, `nbn+jxc+jxt`, `nbn+jxt`, `nbn+nbn`, `nbn+nbn+jp+ef`, `nbn+xsm+ecs`, `nbn+xsm+ef`, `nbn+xsm+ep+ef`, `nbn+xsm+ep+ef+jcr`, `nbn+xsm+etm`, `nbn+xsn`, `nbn+xsn+jca`, `nbn+xsn+jca+jp+ef+jcr`, `nbn+xsn+jca+jxc`, `nbn+xsn+jca+jxt`, `nbn+xsn+jcm`, `nbn+xsn+jco`, `nbn+xsn+jcs`, `nbn+xsn+jct`, `nbn+xsn+jp+ecc`, `nbn+xsn+jp+ecs`, `nbn+xsn+jp+ef`, `nbn+xsn+jp+ef+jcr`, `nbn+xsn+jp+ep+ef`, `nbn+xsn+jxc`, `nbn+xsn+jxt`, `nbn+xsv+etm`, `nbu`, `nbu+jca`, `nbu+jca+jxc`, `nbu+jca+jxt`, `nbu+jcc`, `nbu+jcc+jxc`, `nbu+jcj`, `nbu+jcm`, `nbu+jco`, `nbu+jcs`, `nbu+jct`, `nbu+jct+jxc`, `nbu+jp+ecc`, `nbu+jp+ecs`, `nbu+jp+ef`, `nbu+jp+ef+jcr`, `nbu+jp+ef+jxc`, `nbu+jp+ep+ecc`, `nbu+jp+ep+ecs`, `nbu+jp+ep+ef`, `nbu+jp+ep+ef+jcr`, `nbu+jp+ep+etm`, `nbu+jp+ep+etn+jco`, `nbu+jp+etm`, `nbu+jxc`, `nbu+jxc+jca`, `nbu+jxc+jcs`, `nbu+jxc+jp+ef`, `nbu+jxc+jp+ep+ef`, `nbu+jxc+jxt`, `nbu+jxt`, `nbu+ncn`, `nbu+ncn+jca`, `nbu+ncn+jcm`, `nbu+xsn`, `nbu+xsn+jca`, `nbu+xsn+jca+jxc`, `nbu+xsn+jca+jxt`, `nbu+xsn+jcm`, `nbu+xsn+jco`, `nbu+xsn+jcs`, `nbu+xsn+jp+ecs`, `nbu+xsn+jp+ep+ef`, `nbu+xsn+jxc`, `nbu+xsn+jxc+jxt`, `nbu+xsn+jxt`, `nbu+xsv+ecc`, `nbu+xsv+etm`, `ncn`, `ncn+f+ncpa+jco`, `ncn+jca`, `ncn+jca+jca`, `ncn+jca+jcc`, `ncn+jca+jcj`, `ncn+jca+jcm`, `ncn+jca+jcs`, `ncn+jca+jct`, `ncn+jca+jp+ecc`, `ncn+jca+jp+ecs`, `ncn+jca+jp+ef`, `ncn+jca+jp+ep+ef`, `ncn+jca+jp+etm`, `ncn+jca+jp+etn+jxt`, `ncn+jca+jxc`, `ncn+jca+jxc+jcc`, `ncn+jca+jxc+jcm`, `ncn+jca+jxc+jxc`, `ncn+jca+jxc+jxt`, `ncn+jca+jxt`, `ncn+jcc`, `ncn+jcc+jxc`, `ncn+jcj`, `ncn+jcj+jxt`, `ncn+jcm`, `ncn+jco`, `ncn+jcr`, `ncn+jcr+jxc`, `ncn+jcs`, `ncn+jcs+jxt`, `ncn+jct`, `ncn+jct+jcm`, `ncn+jct+jxc`, `ncn+jct+jxt`, `ncn+jcv`, `ncn+jp+ecc`, `ncn+jp+ecc+jct`, `ncn+jp+ecc+jxc`, `ncn+jp+ecs`, `ncn+jp+ecs+jcm`, `ncn+jp+ecs+jco`, `ncn+jp+ecs+jxc`, `ncn+jp+ecs+jxt`, `ncn+jp+ecx`, `ncn+jp+ef`, `ncn+jp+ef+jca`, `ncn+jp+ef+jcm`, `ncn+jp+ef+jco`, `ncn+jp+ef+jcr`, `ncn+jp+ef+jcr+jxc`, `ncn+jp+ef+jcr+jxt`, `ncn+jp+ef+jp+etm`, `ncn+jp+ef+jxc`, `ncn+jp+ef+jxf`, `ncn+jp+ef+jxt`, `ncn+jp+ep+ecc`, `ncn+jp+ep+ecs`, `ncn+jp+ep+ecs+jxc`, `ncn+jp+ep+ecx`, `ncn+jp+ep+ef`, `ncn+jp+ep+ef+jcr`, `ncn+jp+ep+ef+jcr+jxc`, `ncn+jp+ep+ef+jxc`, `ncn+jp+ep+ef+jxf`, `ncn+jp+ep+ef+jxt`, `ncn+jp+ep+ep+etm`, `ncn+jp+ep+etm`, `ncn+jp+ep+etn`, `ncn+jp+ep+etn+jca`, `ncn+jp+ep+etn+jca+jxc`, `ncn+jp+ep+etn+jco`, `ncn+jp+ep+etn+jcs`, `ncn+jp+ep+etn+jxt`, `ncn+jp+etm`, `ncn+jp+etn`, `ncn+jp+etn+jca`, `ncn+jp+etn+jca+jxc`, `ncn+jp+etn+jca+jxt`, `ncn+jp+etn+jco`, `ncn+jp+etn+jcs`, `ncn+jp+etn+jct`, `ncn+jp+etn+jxc`, `ncn+jp+etn+jxt`, `ncn+jxc`, `ncn+jxc+jca`, `ncn+jxc+jca+jxc`, `ncn+jxc+jca+jxt`, `ncn+jxc+jcc`, `ncn+jxc+jcm`, `ncn+jxc+jco`, `ncn+jxc+jcs`, `ncn+jxc+jct+jxt`, `ncn+jxc+jp+ef`, `ncn+jxc+jp+ef+jcr`, `ncn+jxc+jp+ep+ecs`, `ncn+jxc+jp+ep+ef`, `ncn+jxc+jp+etm`, `ncn+jxc+jxc`, `ncn+jxc+jxt`, `ncn+jxt`, `ncn+jxt+jcm`, `ncn+jxt+jxc`, `ncn+nbn`, `ncn+nbn+jca`, `ncn+nbn+jcm`, `ncn+nbn+jcs`, `ncn+nbn+jp+ecc`, `ncn+nbn+jp+ep+ef`, `ncn+nbn+jxc`, `ncn+nbn+jxt`, `ncn+nbu`, `ncn+nbu+jca`, `ncn+nbu+jcm`, `ncn+nbu+jco`, `ncn+nbu+jp+ef`, `ncn+nbu+jxc`, `ncn+nbu+ncn`, `ncn+ncn`, `ncn+ncn+jca`, `ncn+ncn+jca+jcc`, `ncn+ncn+jca+jcm`, `ncn+ncn+jca+jxc`, `ncn+ncn+jca+jxc+jcm`, `ncn+ncn+jca+jxc+jxc`, `ncn+ncn+jca+jxt`, `ncn+ncn+jcc`, `ncn+ncn+jcj`, `ncn+ncn+jcm`, `ncn+ncn+jco`, `ncn+ncn+jcr`, `ncn+ncn+jcs`, `ncn+ncn+jct`, `ncn+ncn+jct+jcm`, `ncn+ncn+jct+jxc`, `ncn+ncn+jct+jxt`, `ncn+ncn+jp+ecc`, `ncn+ncn+jp+ecs`, `ncn+ncn+jp+ef`, `ncn+ncn+jp+ef+jcm`, `ncn+ncn+jp+ef+jcr`, `ncn+ncn+jp+ef+jcs`, `ncn+ncn+jp+ep+ecc`, `ncn+ncn+jp+ep+ecs`, `ncn+ncn+jp+ep+ef`, `ncn+ncn+jp+ep+ef+jcr`, `ncn+ncn+jp+ep+ep+etm`, `ncn+ncn+jp+ep+etm`, `ncn+ncn+jp+ep+etn`, `ncn+ncn+jp+etm`, `ncn+ncn+jp+etn`, `ncn+ncn+jp+etn+jca`, `ncn+ncn+jp+etn+jco`, `ncn+ncn+jp+etn+jxc`, `ncn+ncn+jxc`, `ncn+ncn+jxc+jca`, `ncn+ncn+jxc+jcc`, `ncn+ncn+jxc+jcm`, `ncn+ncn+jxc+jco`, `ncn+ncn+jxc+jcs`, `ncn+ncn+jxc+jxc`, `ncn+ncn+jxt`, `ncn+ncn+nbn`, `ncn+ncn+ncn`, `ncn+ncn+ncn+jca`, `ncn+ncn+ncn+jca+jcm`, `ncn+ncn+ncn+jca+jxt`, `ncn+ncn+ncn+jcj`, `ncn+ncn+ncn+jcm`, `ncn+ncn+ncn+jco`, `ncn+ncn+ncn+jcs`, `ncn+ncn+ncn+jct+jxt`, `ncn+ncn+ncn+jp+etn+jxc`, `ncn+ncn+ncn+jxt`, `ncn+ncn+ncn+ncn+jca`, `ncn+ncn+ncn+ncn+jca+jxt`, `ncn+ncn+ncn+ncn+jco`, `ncn+ncn+ncn+xsn+jp+etm`, `ncn+ncn+ncpa`, `ncn+ncn+ncpa+jca`, `ncn+ncn+ncpa+jcm`, `ncn+ncn+ncpa+jco`, `ncn+ncn+ncpa+jcs`, `ncn+ncn+ncpa+jxc`, `ncn+ncn+ncpa+jxt`, `ncn+ncn+ncpa+ncn`, `ncn+ncn+ncpa+ncn+jca`, `ncn+ncn+ncpa+ncn+jcj`, `ncn+ncn+ncpa+ncn+jcm`, `ncn+ncn+ncpa+ncn+jxt`, `ncn+ncn+xsn`, `ncn+ncn+xsn+jca`, `ncn+ncn+xsn+jca+jxt`, `ncn+ncn+xsn+jcj`, `ncn+ncn+xsn+jcm`, `ncn+ncn+xsn+jco`, `ncn+ncn+xsn+jcs`, `ncn+ncn+xsn+jct`, `ncn+ncn+xsn+jp+ecs`, `ncn+ncn+xsn+jp+ep+ef`, `ncn+ncn+xsn+jp+etm`, `ncn+ncn+xsn+jxc`, `ncn+ncn+xsn+jxc+jcs`, `ncn+ncn+xsn+jxt`, `ncn+ncn+xsv+ecc`, `ncn+ncn+xsv+etm`, `ncn+ncpa`, `ncn+ncpa+jca`, `ncn+ncpa+jca+jcm`, `ncn+ncpa+jca+jxc`, `ncn+ncpa+jca+jxt`, `ncn+ncpa+jcc`, `ncn+ncpa+jcj`, `ncn+ncpa+jcm`, `ncn+ncpa+jco`, `ncn+ncpa+jcr`, `ncn+ncpa+jcs`, `ncn+ncpa+jct`, `ncn+ncpa+jct+jcm`, `ncn+ncpa+jct+jxt`, `ncn+ncpa+jp+ecc`, `ncn+ncpa+jp+ecc+jxc`, `ncn+ncpa+jp+ecs`, `ncn+ncpa+jp+ecs+jxc`, `ncn+ncpa+jp+ef`, `ncn+ncpa+jp+ef+jcr`, `ncn+ncpa+jp+ef+jcr+jxc`, `ncn+ncpa+jp+ep+ef`, `ncn+ncpa+jp+ep+etm`, `ncn+ncpa+jp+ep+etn`, `ncn+ncpa+jp+etm`, `ncn+ncpa+jxc`, `ncn+ncpa+jxc+jca+jxc`, `ncn+ncpa+jxc+jco`, `ncn+ncpa+jxc+jcs`, `ncn+ncpa+jxt`, `ncn+ncpa+nbn+jcs`, `ncn+ncpa+ncn`, `ncn+ncpa+ncn+jca`, `ncn+ncpa+ncn+jca+jcm`, `ncn+ncpa+ncn+jca+jxc`, `ncn+ncpa+ncn+jca+jxt`, `ncn+ncpa+ncn+jcj`, `ncn+ncpa+ncn+jcm`, `ncn+ncpa+ncn+jco`, `ncn+ncpa+ncn+jcs`, `ncn+ncpa+ncn+jct`, `ncn+ncpa+ncn+jct+jcm`, `ncn+ncpa+ncn+jp+ef+jcr`, `ncn+ncpa+ncn+jp+ep+etm`, `ncn+ncpa+ncn+jxc`, `ncn+ncpa+ncn+jxt`, `ncn+ncpa+ncn+xsn+jcm`, `ncn+ncpa+ncn+xsn+jxt`, `ncn+ncpa+ncpa`, `ncn+ncpa+ncpa+jca`, `ncn+ncpa+ncpa+jcj`, `ncn+ncpa+ncpa+jcm`, `ncn+ncpa+ncpa+jco`, `ncn+ncpa+ncpa+jcs`, `ncn+ncpa+ncpa+jp+ep+ef`, `ncn+ncpa+ncpa+jxt`, `ncn+ncpa+ncpa+ncn`, `ncn+ncpa+xsn`, `ncn+ncpa+xsn+jcm`, `ncn+ncpa+xsn+jco`, `ncn+ncpa+xsn+jcs`, `ncn+ncpa+xsn+jp+ecc`, `ncn+ncpa+xsn+jp+etm`, `ncn+ncpa+xsn+jxt`, `ncn+ncpa+xsv+ecc`, `ncn+ncpa+xsv+ecs`, `ncn+ncpa+xsv+ecx`, `ncn+ncpa+xsv+ecx+px+etm`, `ncn+ncpa+xsv+ef`, `ncn+ncpa+xsv+ef+jcm`, `ncn+ncpa+xsv+ef+jcr`, `ncn+ncpa+xsv+etm`, (truncated: full list in pipeline meta)
`morphologizer`	`POS=CCONJ`, `POS=ADV`, `POS=SCONJ`, `POS=DET`, `POS=NOUN`, `POS=VERB`, `POS=ADJ`, `POS=PUNCT`, `POS=SPACE`, `POS=AUX`, `POS=PRON`, `POS=PROPN`, `POS=NUM`, `POS=INTJ`, `POS=PART`, `POS=X`, `POS=ADP`, `POS=SYM`
`parser`	`ROOT`, `acl`, `advcl`, `advmod`, `amod`, `appos`, `aux`, `case`, `cc`, `ccomp`, `compound`, `conj`, `cop`, `csubj`, `dep`, `det`, `dislocated`, `fixed`, `flat`, `iobj`, `mark`, `nmod`, `nsubj`, `nummod`, `obj`, `obl`, `punct`, `xcomp`
`ner`	`DT`, `LC`, `OG`, `PS`, `QT`, `TI`

Accuracy

Type	Score
`TOKEN_ACC`	100.00
`TOKEN_P`	100.00
`TOKEN_R`	100.00
`TOKEN_F`	100.00
`TAG_ACC`	84.00
`POS_ACC`	94.88
`SENTS_P`	100.00
`SENTS_R`	100.00
`SENTS_F`	100.00
`DEP_UAS`	84.17
`DEP_LAS`	81.40
`LEMMA_ACC`	90.09
`ENTS_P`	86.69
`ENTS_R`	83.73
`ENTS_F`	85.19

🔧 Technical Details

The model is optimized for CPU and has multiple components for token classification tasks. The following is a summary of the performance metrics for different tasks:

Task	Metric	Value
NER	Precision	0.8669446273
NER	Recall	0.837301307
NER	F Score	0.8518651621
TAG	Accuracy	0.8400253175
POS	Accuracy	0.9487717077
LEMMA	Accuracy	0.9009276291
UNLABELED_DEPENDENCIES	Unlabeled Attachment Score (UAS)	0.8416620252
LABELED_DEPENDENCIES	Labeled Attachment Score (LAS)	0.8140177338
SENTS	F - Score	1.0

📄 License

This project is licensed under the CC BY - SA 4.0 license.

Featured Recommended AI Models

Empowering the Future, Your AI Solution Knowledge Base

English 简体中文繁體中文にほんご