Papers › Improving the previous state-of-the-art Frisian ASR by fine-tuning XLS-R

Improving the previous state-of-the-art Frisian ASR by fine-tuning XLS-R

31 Mar 2023- 2023 3archive 2025-07-28

Dragoș Alexandru Bălan, Golshid Shekoufandeh

Automatic Speech Recognition (ASR), a system that converts human speech to text, has a major role in digitizing human communication. Despite their significance, most of these systems are designed for higher-resourced languages, like English, Mandarin, or Spanish, leaving lower-resourced languages, such as Frisian, underrepresented. To address this issue, our paper introduces a fine-tuned ASR model based on the Wav2Vec 2.0 XLS-R architecture, trained on the Common Voice corpus version 12.0, to transcribe Frisian speech. With a learning rate of 8e-5, our proposed ASR system has achieved a 15.99% word error rate (WER), surpassing the previous state-of-the-art of 16.25% and serving as a benchmark for future research in this field.

PaperPDF

Code

No code repository is listed for this paper in the archive or in Syntology's graph.

Code Syntology ran Syntology

Not run by Syntology. Nothing on this page verifies that the listed code works.

Tasks

Automatic Speech RecognitionAutomatic Speech Recognition (ASR)Speech RecognitionSpeech-to-Textspeech-recognition

Results from the paper archive 2025-07-28

TaskDatasetModelMetricValueRank at snapshotLeaderboardReport
Speech Recognition Common Voice Frisian wav2vec2-large-xls-r-1b-frisian Test WER 15.99% #1 of 1 Archive leaderboard report

Ranks are positions in the archive's leaderboards as they stood at the 2025-07-28 snapshot. Results published since then are not among these rows, so a rank here is not a current standing.

Methods

XLSR

Report a problem or propose a change · a person checks every report against the paper or source before anything changes; decisions are listed on /corrections