TASK	DATASET	MODEL	METRIC NAME	METRIC VALUE	GLOBAL RANK
Czech Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	99.22	# 1
Romanian Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	98.64	# 1
Turkish Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	98.95	# 1
Hungarian Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	99.41	# 1
Croatian Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	99.73	# 1
Spanish Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	99.62	# 1
Irish Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	98.88	# 1
French Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	99.71	# 1
Slovak Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	99.32	# 1
Polish Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	99.66	# 1
Latvian Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	98.63	# 1
Vietnamese Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	98.53	# 1

Badge	Markdown
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/diacritics-restoration-using-bert-with/czech-text-diacritization-on-multilingual)](https://paperswithcode.com/sota/czech-text-diacritization-on-multilingual?p=diacritics-restoration-using-bert-with)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/diacritics-restoration-using-bert-with/romanian-text-diacritization-on-multilingual)](https://paperswithcode.com/sota/romanian-text-diacritization-on-multilingual?p=diacritics-restoration-using-bert-with)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/diacritics-restoration-using-bert-with/turkish-text-diacritization-on-multilingual)](https://paperswithcode.com/sota/turkish-text-diacritization-on-multilingual?p=diacritics-restoration-using-bert-with)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/diacritics-restoration-using-bert-with/hungarian-text-diacritization-on-multilingual)](https://paperswithcode.com/sota/hungarian-text-diacritization-on-multilingual?p=diacritics-restoration-using-bert-with)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/diacritics-restoration-using-bert-with/croatian-text-diacritization-on-multilingual)](https://paperswithcode.com/sota/croatian-text-diacritization-on-multilingual?p=diacritics-restoration-using-bert-with)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/diacritics-restoration-using-bert-with/spanish-text-diacritization-on-multilingual)](https://paperswithcode.com/sota/spanish-text-diacritization-on-multilingual?p=diacritics-restoration-using-bert-with)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/diacritics-restoration-using-bert-with/irish-text-diacritization-on-multilingual)](https://paperswithcode.com/sota/irish-text-diacritization-on-multilingual?p=diacritics-restoration-using-bert-with)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/diacritics-restoration-using-bert-with/french-text-diacritization-on-multilingual)](https://paperswithcode.com/sota/french-text-diacritization-on-multilingual?p=diacritics-restoration-using-bert-with)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/diacritics-restoration-using-bert-with/slovak-text-diacritization-on-multilingual)](https://paperswithcode.com/sota/slovak-text-diacritization-on-multilingual?p=diacritics-restoration-using-bert-with)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/diacritics-restoration-using-bert-with/polish-text-diacritization-on-multilingual)](https://paperswithcode.com/sota/polish-text-diacritization-on-multilingual?p=diacritics-restoration-using-bert-with)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/diacritics-restoration-using-bert-with/latvian-text-diacritization-on-multilingual)](https://paperswithcode.com/sota/latvian-text-diacritization-on-multilingual?p=diacritics-restoration-using-bert-with)`
	`[![PWC](https://img.shields.io/endpoint.svg?url=https://paperswithcode.com/badge/diacritics-restoration-using-bert-with/vietnamese-text-diacritization-on)](https://paperswithcode.com/sota/vietnamese-text-diacritization-on?p=diacritics-restoration-using-bert-with)`

Diacritics Restoration using BERT with Analysis on Czech language

24 May 2021 · Jakub Náplava, Milan Straka, Jana Straková ·

We propose a new architecture for diacritics restoration based on contextualized embeddings, namely BERT, and we evaluate it on 12 languages with diacritics. Furthermore, we conduct a detailed error analysis on Czech, a morphologically rich language with a high level of diacritization. Notably, we manually annotate all mispredictions, showing that roughly 44% of them are actually not errors, but either plausible variants (19%), or the system corrections of erroneous data (25%). Finally, we categorize the real errors in detail. We release the code at https://github.com/ufal/bert-diacritics-restoration.

PDF Abstract

Code

Add Remove Mark official

ufal/bert-diacritics-restoration official

Tasks

Add Remove

Croatian Text Diacritization

Czech Text Diacritization

French Text Diacritization

Hungarian Text Diacritization

Irish Text Diacritization

Latvian Text Diacritization

Polish Text Diacritization

Romanian Text Diacritization

Slovak Text Diacritization

Spanish Text Diacritization

Turkish Text Diacritization

Vietnamese Text Diacritization

Datasets

Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems

Results from the Paper

Edit

Ranked #1 on Czech Text Diacritization on Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems

Get a GitHub badge

Task	Dataset	Model	Metric Name	Metric Value	Global Rank	Benchmark
Czech Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	99.22	# 1	Compare
Romanian Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	98.64	# 1	Compare
Turkish Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	98.95	# 1	Compare
Hungarian Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	99.41	# 1	Compare
Croatian Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	99.73	# 1	Compare
Spanish Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	99.62	# 1	Compare
Irish Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	98.88	# 1	Compare
French Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	99.71	# 1	Compare
Slovak Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	99.32	# 1	Compare
Polish Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	99.66	# 1	Compare
Latvian Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	98.63	# 1	Compare
Vietnamese Text Diacritization	Multilingual Dataset for Training and Evaluating Diacritics Restoration Systems	BERT	Alpha-Word accuracy	98.53	# 1	Compare

Methods

Add Remove

Adam • Attention Dropout • BERT • Dense Connections • Dropout • GELU • Layer Normalization • Linear Layer • Linear Warmup With Linear Decay • Multi-Head Attention • Residual Connection • Scaled Dot-Product Attention • Softmax • Weight Decay • WordPiece

Edit Social Preview

Diacritics Restoration using BERT with Analysis on Czech language

Code Edit Add Remove Mark official

Tasks Edit Add Remove

Datasets Edit

Results from the Paper Edit

Methods Edit Add Remove

Code

Add Remove Mark official

Tasks

Add Remove

Datasets

Results from the Paper

Edit

Methods

Add Remove