The open-source Icelandic Part-of-Speech tagging model, xlm-roberta-base-ft-udpos28-is

Xlm Roberta Base Ft Udpos28 Is

Developed by wietsedv

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

Sequence Labeling

Transformers

OtherOpen Source License:Apache-2.0 #Icelandic POS tagging #Multilingual transfer learning #High accuracy (95.6%)

Downloads 15

Release Time : 3/2/2022

Model Overview

This model is specifically designed for Icelandic POS tagging tasks, capable of identifying and labeling the part-of-speech categories for each word in the text.

Model Features

Multilingual Support

Based on the XLM-RoBERTa architecture, it has cross-lingual transfer capabilities.

High Accuracy

Achieves 95.6% accuracy on the Icelandic test set.

Universal Dependencies Standard

Follows the POS tagging standard of the Universal Dependencies dataset v2.8.

Model Capabilities

POS Tagging

Text Analysis

Language Processing

Use Cases

Natural Language Processing

Icelandic Text Analysis

Performs POS tagging on Icelandic text to support downstream NLP tasks.

Accuracy reaches 95.6%

Linguistic Research

Used for Icelandic grammar analysis and linguistic feature studies.

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

This model is designed for part - of - speech tagging in Icelandic, leveraging the XLM - RoBERTa base architecture on Universal Dependencies v2.8 dataset. It provides a solution for cross - lingual transfer in POS tagging across over 100 languages.

🚀 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.

✨ Features

Multilingual Support: Capable of part - of - speech tagging in over 100 languages.
High Accuracy: Demonstrates high accuracy in various languages as shown in the performance metrics.

💻 Usage Examples

Basic Usage

from transformers import AutoTokenizer, AutoModelForTokenClassification

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

📚 Documentation

Model Information

Property	Details
Model Type	xlm - roberta - base - ft - udpos28 - is
Training Data	Universal Dependencies v2.8
Library Name	transformers
Tags	part - of - speech, token - classification

Performance Metrics

The model has been evaluated on multiple languages with the following accuracy results:

Language	Test Accuracy
English	88.4
Dutch	86.9
German	82.7
Italian	84.6
French	83.6
Spanish	83.6
Russian	87.6
Swedish	89.9
Norwegian	86.4
Danish	89.6
Low Saxon	57.6
Akkadian	30.5
Armenian	86.6
Welsh	66.9
Old East Slavic	76.3
Albanian	80.8
Slovenian	76.8
Guajajara	31.8
Kurmanji	78.6
Turkish	77.3
Finnish	84.8
Indonesian	84.4
Ukrainian	85.9
Polish	84.2
Portuguese	86.6
Kazakh	81.8
Latin	75.8
Old French	58.6
Buryat	63.1
Kaapor	18.3
Korean	64.3
Estonian	86.7
Croatian	86.0
Gothic	26.6
Swiss German	45.6
Assyrian	15.5
North Sami	43.9
Naija	46.6
Latvian	85.3
Chinese	60.4
Tagalog	80.0
Bambara	32.5
Lithuanian	85.9
Galician	80.7
Vietnamese	64.1
Greek	80.5
Catalan	82.7
Czech	84.6
Erzya	52.8
Bhojpuri	59.0
Thai	68.2
Marathi	87.1
Basque	79.5
Slovak	86.0
Kiche	42.2
Yoruba	34.3
Warlpiri	43.7
Tamil	83.9
Maltese	27.5
Ancient Greek	64.0
Icelandic	95.6
Mbya Guarani	31.9
Urdu	72.7
Romanian	82.0
Persian	78.3
Apurina	47.9
Japanese	44.0
Hungarian	77.2
Hindi	77.4
Classical Chinese	46.0
Komi Permyak	52.7
Faroese	83.9
Sanskrit	37.4
Livvi	66.8
Arabic	79.2
Wolof	39.9
Bulgarian	87.7
Akuntsu	37.0
Makurap	24.7
Kangri	50.2
Breton	61.8
Telugu	84.5
Cantonese	60.6
Old Church Slavonic	53.9
Karelian	74.0
Upper Sorbian	75.5
South Levantine Arabic	70.8
Komi Zyrian	47.1
Irish	66.8
Nayini	43.6
Munduruku	28.3
Manx	48.6
Skolt Sami	39.6
Afrikaans	87.4
Old Turkish	38.9
Tupinamba	37.6
Belarusian	86.8
Serbian	87.2
Moksha	49.8
Western Armenian	79.9
Scottish Gaelic	56.8
Khunsari	52.7
Hebrew	85.4
Uyghur	76.9
Chukchi	37.7

📄 License

This model is released under the Apache 2.0 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