Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Mići Princ -- A Little Boy Teaching Speech Technologies the Chakavian Dialect

Nikola Ljubešić, Peter Rupnik, Tea Perinčić

TL;DR

Mići Princ addresses the lack of open dialectal Croatian resources by releasing a word-aligned Chakavian dataset derived from a translated classic. The authors construct two data flavours (verbatim MP.json and ASR-focused MP.asr.json) from careful diarisation, alignment, and normalization, and publish it in CLARIN.SI and HuggingFace. They adapt the Whisper-large-v3 ASR model to Chakavian with 80 epochs, achieving substantial reductions in word and character error rates (WER ≈ 16.83%, CER ≈ 3.95%). The results demonstrate that even small dialectal corpora can meaningfully improve ASR when combined with robust preprocessing and alignment. The work also paves the way for digital online editions and broader dialectal linguistic research.

Abstract

This paper documents our efforts in releasing the printed and audio book of the translation of the famous novel The Little Prince into the Chakavian dialect, as a computer-readable, AI-ready dataset, with the textual and the audio components of the two releases now aligned on the level of each written and spoken word. Our motivation for working on this release is multiple. The first one is our wish to preserve the highly valuable and specific content beyond the small editions of the printed and the audio book. With the dataset published in the CLARIN.SI repository, this content is from now on at the fingertips of any interested individual. The second motivation is to make the data available for various artificial-intelligence-related usage scenarios, such as the one we follow upon inside this paper already -- adapting the Whisper-large-v3 open automatic speech recognition model, with decent performance on standard Croatian, to Chakavian dialectal speech. We can happily report that with adapting the model, the word error rate on the selected test data has being reduced to a half, while we managed to remove up to two thirds of the error on character level. We envision many more usages of this dataset beyond the set of experiments we have already performed, both on tasks of artificial intelligence research and application, as well as dialectal research. The third motivation for this release is our hope that this, now highly structured dataset, will be transformed into a digital online edition of this work, allowing individuals beyond the research and technology communities to enjoy the beauty of the message of the little boy in the desert, told through the spectacular prism of the Chakavian dialect.

Mići Princ -- A Little Boy Teaching Speech Technologies the Chakavian Dialect

TL;DR

Mići Princ addresses the lack of open dialectal Croatian resources by releasing a word-aligned Chakavian dataset derived from a translated classic. The authors construct two data flavours (verbatim MP.json and ASR-focused MP.asr.json) from careful diarisation, alignment, and normalization, and publish it in CLARIN.SI and HuggingFace. They adapt the Whisper-large-v3 ASR model to Chakavian with 80 epochs, achieving substantial reductions in word and character error rates (WER ≈ 16.83%, CER ≈ 3.95%). The results demonstrate that even small dialectal corpora can meaningfully improve ASR when combined with robust preprocessing and alignment. The work also paves the way for digital online editions and broader dialectal linguistic research.

Abstract

This paper documents our efforts in releasing the printed and audio book of the translation of the famous novel The Little Prince into the Chakavian dialect, as a computer-readable, AI-ready dataset, with the textual and the audio components of the two releases now aligned on the level of each written and spoken word. Our motivation for working on this release is multiple. The first one is our wish to preserve the highly valuable and specific content beyond the small editions of the printed and the audio book. With the dataset published in the CLARIN.SI repository, this content is from now on at the fingertips of any interested individual. The second motivation is to make the data available for various artificial-intelligence-related usage scenarios, such as the one we follow upon inside this paper already -- adapting the Whisper-large-v3 open automatic speech recognition model, with decent performance on standard Croatian, to Chakavian dialectal speech. We can happily report that with adapting the model, the word error rate on the selected test data has being reduced to a half, while we managed to remove up to two thirds of the error on character level. We envision many more usages of this dataset beyond the set of experiments we have already performed, both on tasks of artificial intelligence research and application, as well as dialectal research. The third motivation for this release is our hope that this, now highly structured dataset, will be transformed into a digital online edition of this work, allowing individuals beyond the research and technology communities to enjoy the beauty of the message of the little boy in the desert, told through the spectacular prism of the Chakavian dialect.
Paper Structure (12 sections, 2 figures, 1 table)

This paper contains 12 sections, 2 figures, 1 table.

Table of Contents

  1. Introduction
  2. Origin of the Data -- The Mići Princ Book
  3. Description of the data processing pipeline
  4. Final encoding of the resulting dataset
  5. MP.json encoding
  6. MP.asr.json encoding
  7. ASR experiments
  8. Finetuning setup
  9. Evaluation
  10. Speaker-based error analysis
  11. Limitations
  12. Conclusions

Figures (2)

  • Figure 1: Origins of the voice actors, plotted on a map of northern Kvarner Gulf (Croatia). Speakers with labelled markers were used for model evaluation (see Section \ref{['sec:asr_experiments']}).
  • Figure 2: Metrics, achieved by the model during fine-tuning.