Faculty of Science, Queensland University of Technology (QUT), Brisbane, Queensland, Australia.
PLoS One. 2021 May 12;16(5):e0250363. doi: 10.1371/journal.pone.0250363. eCollection 2021.
Bird call libraries are difficult to collect yet vital for bio-acoustics studies. A potential solution is citizen science labelling of calls. However, acoustic annotation techniques are still relatively undeveloped and in parallel, citizen science initiatives struggle with maintaining participant engagement, while increasing efficiency and accuracy. This study explores the use of an under-utilised and theoretically engaging and intuitive means of sound categorisation: onomatopoeia. To learn if onomatopoeia was a reliable means of categorisation, an online experiment was conducted. Participants sourced from Amazon mTurk (N = 104) ranked how well twelve onomatopoeic words described acoustic recordings of ten native Australian bird calls. Of the ten bird calls, repeated measures ANOVA revealed that five of these had single descriptors ranked significantly higher than all others, while the remaining calls had multiple descriptors that were rated significantly higher than others. Agreement as assessed by Kendall's W shows that overall, raters agreed regarding the suitability and unsuitability of the descriptors used across all bird calls. Further analysis of the spread of responses using frequency charts confirms this and indicates that agreement on which descriptors were unsuitable was pronounced throughout, and that stronger agreement of suitable singular descriptions was matched with greater rater confidence. This demonstrates that onomatopoeia may be reliably used to classify bird calls by non-expert listeners, adding to the suite of methods used in classification of biological sounds. Interface design implications for acoustic annotation are discussed.
鸟鸣库难以收集，但对于生物声学研究至关重要。一种潜在的解决方案是公民科学对鸟鸣进行标记。然而，声学注释技术仍相对不成熟，同时，公民科学计划在保持参与者参与度的同时，还在努力提高效率和准确性。本研究探索了一种利用声音分类的未充分利用且具有理论吸引力和直观性的方法：拟声词。为了了解拟声词是否是一种可靠的分类方法，进行了一项在线实验。参与者来自亚马逊 mTurk（N = 104），他们对 12 个拟声词描述 10 种澳大利亚本地鸟类叫声的录音的程度进行了排名。在这 10 种鸟鸣中，重复测量方差分析显示，其中 5 种鸟鸣的单一描述符评分显著高于其他所有描述符，而其余鸟鸣有多个描述符的评分显著高于其他描述符。肯德尔 W 检验评估的一致性表明，总体而言，评分者在所有鸟鸣的描述符的适用性和不适用性方面达成一致意见。使用频率图表进一步分析响应的分布情况证实了这一点，并表明对于哪些描述符不适用的一致性非常明显，而对于合适的单一描述符的更强一致性则与更高的评分者信心相匹配。这表明，非专业听众可以使用拟声词可靠地对鸟鸣进行分类，这增加了用于生物声音分类的方法套件。讨论了对声学注释的界面设计的影响。
