The la_core_web_lg open-source Latin language processing model - Supports a variety of natural language processing tasks

La Core Web Lg

Developed by latincy

A Latin processing model based on spaCy, supporting multiple natural language processing tasks.

Sequence Labeling OtherOpen Source License:MIT #Latin NLP #Multi-task analysis #High-precision morphological analysis

Downloads 272

Release Time : 4/29/2023

Model Overview

la_core_web_lg is an efficient Latin processing model capable of performing multiple natural language processing tasks such as named entity recognition, part-of-speech tagging, and morphological analysis.

Model Features

Multi-task support

Can handle multiple natural language processing tasks such as NER, TAG, POS, MORPH, LEMMA, and Dependency Analysis simultaneously

High accuracy

Achieved high scores on various task evaluation metrics, such as an NER F Score of 0.859

Comprehensive training data

Trained on multiple Latin corpora, including UD_Latin and LASLA Corpus

Model Capabilities

Named Entity Recognition

Part-of-Speech Tagging

Morphological Analysis

Dependency Analysis

Sentence Segmentation

Lemmatization

Use Cases

Classical literature research

Latin literature analysis

Used to analyze the grammatical structure and entity recognition in classical Latin literature

Can accurately identify entities such as person names and place names in the literature

Language education

Latin learning assistance

Helps students understand the morphological changes and grammatical structure of Latin vocabulary

Provides accurate part-of-speech tagging and morphological analysis

🚀 la_core_web_lg

This is a spaCy model for Latin language processing, capable of performing various token - classification tasks such as NER, POS tagging, etc.

📚 Documentation

Model Information

Property	Details
Model Name	`la_core_web_lg`
Version	`3.8.0`
spaCy Compatibility	`>=3.8.3,<3.9.0`
Default Pipeline	`senter`, `normer`, `tok2vec`, `tagger`, `morphologizer`, `trainable_lemmatizer`, `parser`, `lookup_lemmatizer`, `ner`
Components	`senter`, `normer`, `tok2vec`, `tagger`, `morphologizer`, `trainable_lemmatizer`, `parser`, `lookup_lemmatizer`, `ner`
Vectors	-1 keys, 200000 unique vectors (300 dimensions)
Sources	UD_Latin - Perseus (via Gamba/Zeman 2023) UD_Latin - PROIEL (via Gamba/Zeman 2023) UD_Latin - ITTB (via Gamba/Zeman 2023) UD_Latin - LLCT (via Gamba/Zeman 2023 UD_Latin - UDante (via Gamba/Zeman 2023) CIRCSE/LASLA: LASLA Corpus UD_Latin - CIRCSE LatinCy Assets
License	`MIT`
Author	Patrick J. Burns; with Nora Bernhardt [ner], Tim Geelhaar [tagger, morphologizer, parser, ner], Vincent Koch [ner]

Performance Metrics

The model has the following performance metrics for different token - classification tasks:

Task	Metric	Value
NER	NER Precision	0.8229665072
NER	NER Recall	0.8986122449
NER	NER F Score	0.8591274487
TAG	TAG (XPOS) Accuracy	0.9542970414
POS	POS (UPOS) Accuracy	0.9726288589
MORPH	Morph (UFeats) Accuracy	0.9258321497
LEMMA	Lemma Accuracy	0.9486704917
UNLABELED_DEPENDENCIES	Unlabeled Attachment Score (UAS)	0.8402730813
LABELED_DEPENDENCIES	Labeled Attachment Score (LAS)	0.7889132477
SENTS	Sentences F - Score	0.9915634929

Label Scheme

View label scheme (1055 labels for 4 components)

| Component | Labels | | --- | --- | | **`tagger`** | `ADJ`, `ADP`, `ADV`, `CCONJ`, `DET`, `INTJ`, `NOUN`, `PART`, `SCONJ`, `X`, `_`, `adjective`, `adverb`, `conjunction`, `determiner`, `interjection`, `noun`, `number`, `particle`, `preposition`, `pronoun`, `proper_noun`, `punc`, `unknown`, `verb` | | **`morphologizer`** | A large number of morphological labels such as `POS=ADV`, `Case=Nom\|Gender=Masc\|Number=Sing\|POS=NOUN`, etc. (The full list is very long and is provided in the original document) |

📄 License

This project is licensed under the MIT 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