Finding Meanings in Low Dimensional Structures: Stochastic Neighbor Embedding Applied to the Analysis of Indri indri Vocal Repertoire
, Chiara De Gregorio 1, Valeria Torti 1, Longondraza Miaretsoa 1, Olivier Friard 1, Rose Marie Randrianarison 2,3, Cristina Giacoma 1 and Marco Gamba 1Round 1
Reviewer 1 Report
This is a well-written, well-organized and well-illustrated paper. It presents the results of original research which deserves to be published.
I have only a minor comment to improve the quality of the work, it would be useful to provide also in the supplementary materials, figures where the spectrograms of the individual vocalizations are better visible.
Author Response
We are grateful to Reviewer 1 for the valuable comments and suggestions. As advised by Reviewer 1, we added two figures better showing the spectrographic features of vocal types in the supplementary materials.
Reviewer 2 Report
See attached file
Comments for author File: 
Comments.pdf
Author Response
We are grateful to Reviewer 2 for the precious comments and suggestions. Following the indications, we unraveled the confusion about figures by improving their legends.
We have slightly modified the description of the methodology used to extract spectral features, which we think is now easy to understand and fully replicable.
We also compared the result of the t-SNE approach with that of Principal Component Analyses, and we added the suggested reference, as well as other earlier works.
Finally, we agree with the reviewer about the legitimation of using a human a priori classification of sounds like the gold standard, and we also would like to have some validation on how indris perceive the indicated vocal classes. Unfortunately, there is no information at present regarding the indris’ perceptual abilities. Future studies may involve playback experiments, but while we have been able to conduct playback experiments using the indris’ songs (unpublished data), we have not been able to test the other repertoire’s sounds yet. It is also unclear whether this kind of experiments could help in categorizing the calls, also because the indris do not survive in captivity and playbacks in the wild are always very difficult to perform.
