fi_core_news_lg Open-source Finnish Language Processing Model - Free Deployment, Supports Analysis of Word Positions, Entities, etc.

Fi Core News Lg

Developed by spacy

A CPU-optimized Finnish language processing pipeline provided by spaCy, featuring comprehensive NLP capabilities including POS tagging, dependency parsing, and named entity recognition

Sequence Labeling Other#Finnish NLP #Multitasking #High-accuracy POS tagging

Downloads 53

Release Time : 5/2/2022

Model Overview

This is a large-scale Finnish natural language processing model with a complete text processing pipeline, supporting tasks such as POS tagging, morphological analysis, dependency parsing, and named entity recognition. The model is optimized for CPU usage and includes a trainable lemmatizer.

Model Features

CPU optimization

Specifically optimized for CPU usage scenarios, suitable for deployment in environments without GPUs

Complete NLP pipeline

Provides a full natural language processing pipeline from tokenization to named entity recognition

High-accuracy POS tagging

POS tagging accuracy reaches 97.09% (XPOS) and 96.28% (UPOS)

Trainable lemmatizer

Includes a trainable lemmatization component with an accuracy of 86.53%

Model Capabilities

POS tagging

Morphological analysis

Dependency parsing

Named entity recognition

Lemmatization

Sentence segmentation

Use Cases

Text analysis

Finnish text grammatical analysis

Performs comprehensive grammatical analysis of Finnish texts, including POS tagging and dependency relation analysis

Unlabeled attachment score (UAS) 83.71%, labeled attachment score (LAS) 79.41%

Finnish named entity recognition

Identifies named entities such as person names, locations, and organizations in Finnish texts

NER F-score 81.83%

Language learning

Finnish learning assistance

Helps learners analyze Finnish sentence structures and lexical morphological changes

Morphological feature accuracy 92.22%

🚀 fi_core_news_lg

A Finnish pipeline optimized for CPU, designed for various token - classification tasks.

📚 Documentation

Details: https://spacy.io/models/fi#fi_core_news_lg

The Finnish pipeline is optimized for CPU. Components include tok2vec, tagger, morphologizer, parser, lemmatizer (trainable_lemmatizer), senter, ner.

Property	Details
Name	`fi_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 Finnish TDT v2.8 (Ginter, Filip; Kanerva, Jenna; Laippala, Veronika; Miekka, Niko; Missilä, Anna; Ojala, Stina; Pyysalo, Sampo) TurkuONE (ffe2040e) (Jouni Luoma, Li - Hsin Chang, Filip Ginter, Sampo Pyysalo) Explosion Vectors (OSCAR 2109 + Wikipedia + OpenSubtitles + WMT News Crawl) (Explosion)
License	`CC BY - SA 4.0`
Author	Explosion

Model Index

The fi_core_news_lg model has been evaluated on several token - classification tasks, and here are the results:

Task	Metric	Value
NER	NER Precision	0.8236272879
NER	NER Recall	0.813030386
NER	NER F Score	0.8182945309
TAG	TAG (XPOS) Accuracy	0.9709439124
POS	POS (UPOS) Accuracy	0.9628474502
MORPH	Morph (UFeats) Accuracy	0.9221890983
LEMMA	Lemma Accuracy	0.8653065672
UNLABELED_DEPENDENCIES	Unlabeled Attachment Score (UAS)	0.8371365653
LABELED_DEPENDENCIES	Labeled Attachment Score (LAS)	0.7941298453
SENTS	Sentences F - Score	0.9083487941

Label Scheme

View label scheme (2145 labels for 4 components)

Component	Labels
`tagger`	`A`, `Adj`, `Adp`, `Adv`, `Adv_V`, `C`, `C_V`, `Foreign`, `Interj`, `N`, `Num`, `Pron`, `Punct`, `Symb`, `V`, `V_Pron`, `_SP`
`morphologizer`	A large number of morphological labels such as `Case=Nom\|Number=Sing\|POS=NOUN`, `NumType=Ord\|POS=ADJ` etc. (The full list is too long to show here, please refer to the original document for details)

📄 License

The model is licensed under CC BY - SA 4.0.

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