# Diacritics Restoration using BERT with Analysis on Czech language

Source: [Diacritics Restoration using BERT with Analysis on Czech language](<https://www.sota2.com/research/paper/68fa6a091e209131cea25a2b>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-01-01T13:26:21.858Z

Authors: Jakub N\\'aplava, Milan Straka, Jana Strakov\\'a

Published: 2021-05-24T16:58:27.000Z

## Abstract

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.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Diacritics restoration on diacritization corpus Czech](<https://www.sota2.com/research/sota/diacritics-restoration-on-diacritization-corpus-czech.md>): Diacritics restoration; diacritization corpus Czech; Alpha-Word Accuracy: not recorded.

- [Diacritics restoration on diacritization corpus Vietnamese](<https://www.sota2.com/research/sota/diacritics-restoration-on-diacritization-corpus-vietnamese.md>): Diacritics restoration; diacritization corpus Vietnamese; Alpha-Word Accuracy: not recorded.

- [Diacritics restoration on diacritization corpus Latvian](<https://www.sota2.com/research/sota/diacritics-restoration-on-diacritization-corpus-latvian.md>): Diacritics restoration; diacritization corpus Latvian; Alpha-Word Accuracy: not recorded.

- [Diacritics restoration on Náplava diacritization corpus Polish](<https://www.sota2.com/research/sota/diacritics-restoration-on-n-plava-diacritization-corpus-polish.md>): Diacritics restoration; Náplava diacritization corpus Polish; Alpha-Word Accuracy: not recorded.

- [Diacritics restoration on diacritization corpus Slovak](<https://www.sota2.com/research/sota/diacritics-restoration-on-diacritization-corpus-slovak.md>): Diacritics restoration; diacritization corpus Slovak; Alpha-Word Accuracy: not recorded.

- [Diacritics restoration on Diacritization Corpus French](<https://www.sota2.com/research/sota/diacritics-restoration-on-diacritization-corpus-french.md>): Diacritics restoration; Diacritization Corpus French; Alpha-Word Accuracy: not recorded.

- [Diacritics restoration on Diacritization corpus Irish](<https://www.sota2.com/research/sota/diacritics-restoration-on-diacritization-corpus-irish.md>): Diacritics restoration; Diacritization corpus Irish; Alpha-Word Accuracy: not recorded.

- [Diacritics restoration on diacritization corpus Spanish](<https://www.sota2.com/research/sota/diacritics-restoration-on-diacritization-corpus-spanish.md>): Diacritics restoration; diacritization corpus Spanish; Alpha-Word Accuracy: not recorded.

- [Diacritics restoration on diacritization corpus Croatian](<https://www.sota2.com/research/sota/diacritics-restoration-on-diacritization-corpus-croatian.md>): Diacritics restoration; diacritization corpus Croatian; Alpha-Word Accuracy: not recorded.

- [Diacritics restoration on Diacritization Corpus Hungarian](<https://www.sota2.com/research/sota/diacritics-restoration-on-diacritization-corpus-hungarian.md>): Diacritics restoration; Diacritization Corpus Hungarian; Alpha-Word Accuracy: not recorded.

- [Diacritics restoration on diacritization corpus Turkish](<https://www.sota2.com/research/sota/diacritics-restoration-on-diacritization-corpus-turkish.md>): Diacritics restoration; diacritization corpus Turkish; Alpha-Word Accuracy: not recorded.

- [Diacritics restoration on diacritization corpus Romanian](<https://www.sota2.com/research/sota/diacritics-restoration-on-diacritization-corpus-romanian.md>): Diacritics restoration; diacritization corpus Romanian; Alpha-Word Accuracy: not recorded.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2105.11408>)

[PDF](<https://arxiv.org/pdf/2105.11408.pdf>)

[Implementation](<https://github.com/ufal/bert-diacritics-restoration>)
