<scp>ArcticBirdSounds</scp>: An open‐access, multiyear, and detailed annotated dataset of bird songs and calls
Bibliographic record
Abstract
Tracking biodiversity shifts is central to understanding past, present, and future global changes. Recent advances in bioacoustics and the low cost of high-quality automatic recorders are revolutionizing studies in biogeography and community and behavioral ecology with a robust assessment of phenology, species occurrence, and individual activity. This large volume of acoustic recordings has recently generated a plethora of datasets that can now be handled automatically, mostly via big data methods such as deep learning. These approaches need high-quality annotations to classify and detect recorded sounds efficiently. However, very few strongly annotated datasets-that is, with detailed information on start and end time of each vocalization-are openly accessible to the public. Moreover, these datasets mostly cover temperate species and are usually limited to a single year of recordings. Here, we present ArcticBirdSounds, the first open-access, multisite, and multiyear strongly annotated dataset of arctic bird vocalizations. ArcticBirdSounds offers 20 h of annotated recordings over 2 years (2018, 2019), taken from 15 distinct plots within six locations across the Arctic, from Alaska to Greenland. Recordings cover the arctic vertebrates' breeding period and are evenly spaced during the day; they capture most species breeding there with 12,933 temporal annotations in 49 classes of sounds. While these data can be used for many pressing ecological questions, it is also a unique resource for methodological development to help meet the challenges of fast ecosystem transformations such as those happening in the Arctic. All data, including audio files, annotation files, and companion spreadsheets, are available in an Open Science Framework repository published under a CC BY 4.0 License.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.001 | 0.003 |
| Meta-epidemiology (narrow) | 0.002 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.002 | 0.003 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.001 | 0.001 |
| Open science | 0.002 | 0.003 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.020 | 0.025 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".