palli23/spjallromur-4h

Dataset Card — Spjallrómur 4h (Icelandic Conversational Speech Test Set)

Dataset Description

Spjallrómur 4h is a curated test set for Automatic Speech Recognition (ASR) evaluation on conversational Icelandic speech. It is a subset of the Spjallrómur 26.03 corpus, compiled specifically for benchmarking ASR systems on spontaneous, dialogue-style Icelandic audio.

The test set was compiled as part of the Almannarómur project: Towards Automatic Speech Recognition for Conversations in Icelandic (Project code: MVF25010163).

• Curated by: Páll Rúnarsson, Research Associate, Reykjavík University
• Funded by: Almannarómur (Language Technology Programme for Icelandic)
• Language(s): Icelandic (is)
• License: CC BY-SA 4.0 — free to use, share, and adapt, provided you give appropriate credit and distribute any derivatives under the same license.
• Source corpus: Spjallrómur 26.03
• Repository: palli23/spjallromur-4h

Dataset Sources

• Source corpus: Spjallrómur 26.03 — CLARIN-IS
• Project: Almannarómur

---

Uses

Direct Use

This dataset is intended for evaluation and benchmarking of Icelandic ASR systems, particularly on conversational and spontaneous speech. It provides a standardised held-out test set for comparing model performance on informal dialogue.

Out-of-Scope Use

• Training or fine-tuning ASR models (this is a test-only split)
• Languages other than Icelandic

---

Dataset Structure

This dataset contains a single test split only. It should not be used for training or hyperparameter tuning, as it is intended solely for final evaluation.

Split	Size
test	~4 hours

Data Fields

• audio — audio recording (16 kHz, mono)
• text (or normalized_text) — reference transcription
• (add any additional fields: speaker ID, turn timestamps, etc.)

Data Collection

Audio and transcriptions are drawn from the Spjallrómur corpus, which contains 54 conversations with 102 speakers, collected by Reykjavík University between September 2020 and September 2021. The 26.03 release includes manual transcriptions with turn-level timestamps and speaker labels.

---

Dataset Creation

Curation Rationale

Existing Icelandic ASR benchmarks focus predominantly on read speech (e.g., Samrómur, Málrómur). This test set was compiled to fill the gap for conversational speech evaluation, enabling fair comparison of ASR systems on spontaneous Icelandic dialogue.

Source Data

Subset of Spjallrómur 26.03, selected and curated by Páll Rúnarsson for use as a standardised test set within the Almannarómur project.

Annotations

Transcriptions are drawn from the manual annotations included in Spjallrómur 26.03, which were produced by three transcribers. Seven conversations include multiple transcriptions for inter-annotator comparison.

---

Bias, Risks, and Limitations

• Speaker demographics reflect those of the Spjallrómur corpus; results may not generalise to all Icelandic speaker populations.
• The test set covers spontaneous conversational speech only; models should not be expected to generalise from read-speech benchmarks.
• As a subset of Spjallrómur, this dataset inherits any biases present in the original collection.

Recommendations

When reporting results on this test set, users should clearly specify the normalisation procedure applied to both reference and hypothesis transcripts to ensure comparability across publications.

---

Citation

If you use this dataset, please cite both this work and the source corpus:

@misc{runarsson2026spjallromur4h,
  author       = {Rúnarsson, Páll},
  title        = {Spjallrómur 4h -- Icelandic Conversational Speech Test Set},
  year         = {2026},
  publisher    = {Hugging Face},
  howpublished = {\url{https://huggingface.co/datasets/palli23/spjallromur-4h}}
}

@misc{fong2026spjallromur,
  author       = {Fong, Judy Y. and Borsky, Michal and Runarsson, Pall
                  and Hedström, Staffan and Jónsson, Ólafur Helgi
                  and Hólmfriðardóttir, Lára Margrét H. and Þorsteinsdóttir, Sunneva
                  and Eiríksdóttir, Málfriður Anna and Mollberg, David Erik
                  and Magnúsdóttir, Eydís Huld and Þórhallsdóttir, Ragnheiður
                  and Gudnason, Jon},
  title        = {Spjallromur 26.03 -- Icelandic Conversational Speech},
  year         = {2026},
  publisher    = {CLARIN-IS / Reykjavík University},
  howpublished = {\url{http://hdl.handle.net/20.500.12537/379}}
}

---

Acknowledgements

This dataset was compiled at the Language and Voice Laboratory (lvl.ru.is) at Reykjavík University, Iceland.

Funded by Almannarómur — the Language Technology Programme for Icelandic, as part of the project Towards Automatic Speech Recognition for Conversations in Icelandic (MVF25010163), funded by the Icelandic Ministry of Education, Science and Culture.

---

Dataset Card Author

Páll Rúnarsson, Language and Voice Laboratory, Reykjavík University

Dataset Card Contact

(add contact email or HF profile link)