The open-source multilingual part-of-speech tagging model, xlm - roberta - base - ft - udpos28 - en, accurately identifies parts of speech in multiple languages.

Xlm Roberta Base Ft Udpos28 En

Developed by wietsedv

This model is a multilingual POS tagging model based on the XLM-RoBERTa architecture, fine-tuned on the Universal Dependencies v2.8 dataset.

Sequence Labeling

Transformers

EnglishOpen Source License:Apache-2.0 #Multilingual POS tagging #High accuracy #Universal Dependencies parsing

Downloads 20

Release Time : 3/2/2022

Model Overview

This is a token classification model for POS tagging tasks, supporting multiple languages with the best performance on English (96% accuracy).

Model Features

Multilingual support

Supports POS tagging tasks for over 100 languages

High accuracy

Achieves 96% accuracy on English test sets

Based on Universal Dependencies dataset

Trained using Universal Dependencies v2.8 dataset

Model Capabilities

POS tagging

Multilingual text processing

Token classification

Use Cases

Natural language processing

Multilingual text analysis

Perform POS tagging on multilingual texts to support syntactic analysis

96% accuracy for English, varying from 90% to 20% for other languages

Linguistic research

Used for comparing grammatical structures across different languages

🚀 XLM-RoBERTa base Universal Dependencies v2.8 POS tagging: English

This model is designed for part - of - speech tagging and is trained on the Universal Dependencies v2.8 dataset. It offers high accuracy across multiple languages, providing a powerful tool for cross - lingual transfer in POS tagging tasks.

🚀 Quick Start

This model is part of our paper called:

Make the Best of Cross - lingual Transfer: Evidence from POS Tagging with over 100 Languages

Check the Space for more details.

💻 Usage Examples

Basic Usage

from transformers import AutoTokenizer, AutoModelForTokenClassification

tokenizer = AutoTokenizer.from_pretrained("wietsedv/xlm-roberta-base-ft-udpos28-en")
model = AutoModelForTokenClassification.from_pretrained("wietsedv/xlm-roberta-base-ft-udpos28-en")

📄 License

The model is licensed under the Apache - 2.0 license.

📚 Documentation

Model Information

Property	Details
Library Name	transformers
Tags	part - of - speech, token - classification
Datasets	universal_dependencies
Metrics	accuracy

Model Results

The model xlm-roberta-base-ft-udpos28-en has the following performance results on different languages in the Part - of - Speech Tagging task using the Universal Dependencies v2.8 dataset:

Language	Test Accuracy
English	96.0
Dutch	90.4
German	88.6
Italian	87.8
French	87.4
Spanish	90.3
Russian	91.0
Swedish	94.0
Norwegian	89.6
Danish	91.6
Low Saxon	57.4
Akkadian	26.4
Armenian	88.5
Welsh	70.6
Old East Slavic	76.5
Albanian	82.3
Slovenian	79.0
Guajajara	17.2
Kurmanji	76.9
Turkish	79.1
Finnish	87.2
Indonesian	86.9
Ukrainian	87.6
Polish	87.2
Portuguese	90.0
Kazakh	82.5
Latin	79.6
Old French	53.4
Buryat	58.8
Kaapor	9.2
Korean	64.0
Estonian	88.4
Croatian	87.9
Gothic	20.5
Swiss German	47.6
Assyrian	14.6
North Sami	32.0
Naija	47.5
Latvian	87.5
Chinese	47.5
Tagalog	73.5
Bambara	27.7
Lithuanian	87.3
Galician	87.1
Vietnamese	66.4
Greek	87.6
Catalan	89.7
Czech	88.1
Erzya	47.6
Bhojpuri	50.7
Thai	59.5
Marathi	82.2
Basque	76.0
Slovak	88.5
Kiche	25.4
Yoruba	18.5
Warlpiri	29.1
Tamil	83.4
Maltese	21.1
Ancient Greek	66.8
Icelandic	84.8
Mbya Guarani	24.1
Urdu	67.0
Romanian	85.7
Persian	76.7
Apurina	28.6
Japanese	34.1
Hungarian	86.0
Hindi	74.1
Classical Chinese	29.4
Komi Permyak	47.4
Faroese	77.0
Sanskrit	25.6
Livvi	63.2
Arabic	80.7
Wolof	26.1
Bulgarian	90.8
Akuntsu	18.3
Makurap	5.5
Kangri	43.0
Breton	64.1
Telugu	84.7
Cantonese	54.0
Old Church Slavonic	53.7
Karelian	69.7
Upper Sorbian	75.6
South Levantine Arabic	66.3
Komi Zyrian	39.9
Irish	67.0
Nayini	44.9
Munduruku	12.3
Manx	25.4
Skolt Sami	29.9
Afrikaans	89.3
Old Turkish	37.1
Tupinamba	23.1
Belarusian	89.1
Serbian	88.4
Moksha	44.1
Western Armenian	80.1
Scottish Gaelic	59.0
Khunsari	43.2
Hebrew	90.6
Uyghur	75.8
Chukchi	32.6

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