Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Killkan: The Automatic Speech Recognition Dataset for Kichwa with Morphosyntactic Information

Chihiro Taguchi, Jefferson Saransig, Dayana Velásquez, David Chiang

TL;DR

Killkan addresses the lack of ASR resources for Kichwa by delivering the first openly available, morphosyntactically annotated ASR dataset built from a public radio corpus. The authors fine-tune wav2vec2-XLSR-53 on Killkan, achieving competitive CERs around $2 ext{-}3 ext{ ext{%}}$ and WERs that improve notably under text normalization, demonstrating viability for extremely small corpora. They also provide UD CoNLL-U annotations with novel morph features for topic/focus and intra-word code-switch boundaries, and analyze morphological complexity and code-switching patterns to inform linguistic and computational modeling. The work advances resource-building for low-resource languages and offers a foundation for future improvements in Kichwa ASR and language technologies, while engaging ethical considerations around accessibility, community needs, and dialectal diversity. Overall, Killkan demonstrates that high-quality ASR and rich linguistic annotation can be achieved for endangered languages with careful data curation, annotation, and analysis, enabling applications and revitalization efforts.

Abstract

This paper presents Killkan, the first dataset for automatic speech recognition (ASR) in the Kichwa language, an indigenous language of Ecuador. Kichwa is an extremely low-resource endangered language, and there have been no resources before Killkan for Kichwa to be incorporated in applications of natural language processing. The dataset contains approximately 4 hours of audio with transcription, translation into Spanish, and morphosyntactic annotation in the format of Universal Dependencies. The audio data was retrieved from a publicly available radio program in Kichwa. This paper also provides corpus-linguistic analyses of the dataset with a special focus on the agglutinative morphology of Kichwa and frequent code-switching with Spanish. The experiments show that the dataset makes it possible to develop the first ASR system for Kichwa with reliable quality despite its small dataset size. This dataset, the ASR model, and the code used to develop them will be publicly available. Thus, our study positively showcases resource building and its applications for low-resource languages and their community.

Killkan: The Automatic Speech Recognition Dataset for Kichwa with Morphosyntactic Information

TL;DR

Killkan addresses the lack of ASR resources for Kichwa by delivering the first openly available, morphosyntactically annotated ASR dataset built from a public radio corpus. The authors fine-tune wav2vec2-XLSR-53 on Killkan, achieving competitive CERs around and WERs that improve notably under text normalization, demonstrating viability for extremely small corpora. They also provide UD CoNLL-U annotations with novel morph features for topic/focus and intra-word code-switch boundaries, and analyze morphological complexity and code-switching patterns to inform linguistic and computational modeling. The work advances resource-building for low-resource languages and offers a foundation for future improvements in Kichwa ASR and language technologies, while engaging ethical considerations around accessibility, community needs, and dialectal diversity. Overall, Killkan demonstrates that high-quality ASR and rich linguistic annotation can be achieved for endangered languages with careful data curation, annotation, and analysis, enabling applications and revitalization efforts.

Abstract

This paper presents Killkan, the first dataset for automatic speech recognition (ASR) in the Kichwa language, an indigenous language of Ecuador. Kichwa is an extremely low-resource endangered language, and there have been no resources before Killkan for Kichwa to be incorporated in applications of natural language processing. The dataset contains approximately 4 hours of audio with transcription, translation into Spanish, and morphosyntactic annotation in the format of Universal Dependencies. The audio data was retrieved from a publicly available radio program in Kichwa. This paper also provides corpus-linguistic analyses of the dataset with a special focus on the agglutinative morphology of Kichwa and frequent code-switching with Spanish. The experiments show that the dataset makes it possible to develop the first ASR system for Kichwa with reliable quality despite its small dataset size. This dataset, the ASR model, and the code used to develop them will be publicly available. Thus, our study positively showcases resource building and its applications for low-resource languages and their community.
Paper Structure (27 sections, 2 equations, 3 figures, 9 tables)

This paper contains 27 sections, 2 equations, 3 figures, 9 tables.

Table of Contents

  1. Introduction
  2. Background
  3. The Kichwa language.
  4. Language contact and code-switching.
  5. Orthography.
  6. Related Work
  7. Dataset
  8. Source
  9. Annotation
  10. Transcription
  11. Translation
  12. Morphosyntactic annotation
  13. Analysis
  14. Morphological complexity.
  15. Code-switching.
  16. ...and 12 more sections

Figures (3)

  • Figure 1: The distribution of Quechua II languages mentioned in this paper. This map was created with the lingtypology library lingtypology.
  • Figure 2: A classification of Quechuan languages with their ISO 639-3 language codes with a focus on the varieties mentioned in this paper. The branches in the box are called Ecuadorian Kichwa.
  • Figure 3: A screenshot of annotating a transcription, its Spanish translation, and UD-style morphosyntactic information in ELAN.