Andrew Mitchell
@mitchellacoustics.bsky.social
Senior Data Scientist, Connected Places Catapult
#OpenScience #Acoustics
drandrewmitchell.com
Host of @JustNoisePod.bsky.social
Developer of Soundscapy
https://soundscapy.readthedocs.io/
#OpenScience #Acoustics
drandrewmitchell.com
Host of @JustNoisePod.bsky.social
Developer of Soundscapy
https://soundscapy.readthedocs.io/
Having a great three days at my favourite sound conference, the Urban Sound Symposium in Zurich!
April 30, 2025 at 9:04 AM
Having a great three days at my favourite sound conference, the Urban Sound Symposium in Zurich!
Our latest article and the capstone paper from the SSID Project!
We've spent 6 years working towards creating useful single-value indices of soundscape quality. We present SPI, our method for defining perception-focussed, context dependent Soundscape Perception Indices!
doi.org/10.1121/10.0...
We've spent 6 years working towards creating useful single-value indices of soundscape quality. We present SPI, our method for defining perception-focussed, context dependent Soundscape Perception Indices!
doi.org/10.1121/10.0...
![Paper Fig 1. Steps for calculating the SPI. (1) Define and parameterise the target circumplex
distribution;
(2) sample the target distribution and prepare the test
distribution;
(3) compare test and target distributions using the distance
metric [two-dimensional Kolmogorov-Smirnov (KS)
distance, DBKS]; and
(4) calculate SPI 100(1 - DBKS)](https://cdn.bsky.app/img/feed_thumbnail/plain/did:plc:rimqaeaslvmfo2bxmdekn2dw/bafkreiftti2e45ix4tw65mwpzulvma7tzampr67g5f6io6xfodugjvwpja@jpeg)
April 18, 2025 at 11:33 AM
Our latest article and the capstone paper from the SSID Project!
We've spent 6 years working towards creating useful single-value indices of soundscape quality. We present SPI, our method for defining perception-focussed, context dependent Soundscape Perception Indices!
doi.org/10.1121/10.0...
We've spent 6 years working towards creating useful single-value indices of soundscape quality. We present SPI, our method for defining perception-focussed, context dependent Soundscape Perception Indices!
doi.org/10.1121/10.0...
There are a ton of views on this, from logical positivism, to critical realism, to falsificationist. Hell, Feyerabend argued for throwing out the scientific method altogether!
November 27, 2024 at 4:40 AM
There are a ton of views on this, from logical positivism, to critical realism, to falsificationist. Hell, Feyerabend argued for throwing out the scientific method altogether!
The developer guidelines for using the AT protocol (which is where I think this dataset creation falls under) state that developers "must have a method for deleting content a user has requested to be deleted"
It is poorly defined, but I think this last point at least was clearly violated by HF.
It is poorly defined, but I think this last point at least was clearly violated by HF.
November 27, 2024 at 3:37 AM
The developer guidelines for using the AT protocol (which is where I think this dataset creation falls under) state that developers "must have a method for deleting content a user has requested to be deleted"
It is poorly defined, but I think this last point at least was clearly violated by HF.
It is poorly defined, but I think this last point at least was clearly violated by HF.
As far as I've found, the most relevant parts are the TOS and developer guidelines. The TOS state that users retain ownership of their content and give permission to BS to use and reproduce it, giving examples of the types of reuse intended.
November 27, 2024 at 3:37 AM
As far as I've found, the most relevant parts are the TOS and developer guidelines. The TOS state that users retain ownership of their content and give permission to BS to use and reproduce it, giving examples of the types of reuse intended.
Whether or not you include more historical context about the concepts you're teaching (and which historical narratives to include) is itself a political decision.
For an example I wrote in response to a CS prof who said he explicitly makes his class apolitical:
For an example I wrote in response to a CS prof who said he explicitly makes his class apolitical:
November 27, 2024 at 3:07 AM
Whether or not you include more historical context about the concepts you're teaching (and which historical narratives to include) is itself a political decision.
For an example I wrote in response to a CS prof who said he explicitly makes his class apolitical:
For an example I wrote in response to a CS prof who said he explicitly makes his class apolitical:
I'm not certain it violates GDPR, but I am more convinced it potentially violates both Bluesky & Hugging Face TOS.
BS says authors retain ownership & HF says you must have explicit consent. Just because content is public and accessible via an open API does not void license and copyright protections
BS says authors retain ownership & HF says you must have explicit consent. Just because content is public and accessible via an open API does not void license and copyright protections
November 27, 2024 at 2:41 AM
I'm not certain it violates GDPR, but I am more convinced it potentially violates both Bluesky & Hugging Face TOS.
BS says authors retain ownership & HF says you must have explicit consent. Just because content is public and accessible via an open API does not void license and copyright protections
BS says authors retain ownership & HF says you must have explicit consent. Just because content is public and accessible via an open API does not void license and copyright protections
It also potentially breaks Hugging Face's own TOS as well, since they did not obtain *explicit* consent. Just because content is public and accessible via an open API does not void license and copyright protections.
November 27, 2024 at 2:15 AM
It also potentially breaks Hugging Face's own TOS as well, since they did not obtain *explicit* consent. Just because content is public and accessible via an open API does not void license and copyright protections.
Back at the British Medical Association for day two of a great seminar on Open Source, Ethics, & Innovation in AI with @turingway.bsky.social #PractitionersHub!
#OSEIAI24
#OSEIAI24
November 26, 2024 at 12:08 PM
Back at the British Medical Association for day two of a great seminar on Open Source, Ethics, & Innovation in AI with @turingway.bsky.social #PractitionersHub!
#OSEIAI24
#OSEIAI24
Public. Data. Repositories.
Research data should be uploaded to data repos where researchers can't (understandably) lose it and which don't have an incentive to magically misplace it (like journals, funders, and unis)
Zenodo, OSF, & figshare all have private options so embargo isn't a valid excuse
Research data should be uploaded to data repos where researchers can't (understandably) lose it and which don't have an incentive to magically misplace it (like journals, funders, and unis)
Zenodo, OSF, & figshare all have private options so embargo isn't a valid excuse
November 20, 2024 at 8:55 PM
Public. Data. Repositories.
Research data should be uploaded to data repos where researchers can't (understandably) lose it and which don't have an incentive to magically misplace it (like journals, funders, and unis)
Zenodo, OSF, & figshare all have private options so embargo isn't a valid excuse
Research data should be uploaded to data repos where researchers can't (understandably) lose it and which don't have an incentive to magically misplace it (like journals, funders, and unis)
Zenodo, OSF, & figshare all have private options so embargo isn't a valid excuse
Public. Data. Repositories.
Research should be uploaded to data repos where researchers can't (understandably) lose it and which have no incentive to magically make it disappear (like funders, journals, or unis).
Zenodo, OSF, figshare all have private options, so embargo isn't a valid excuse.
Research should be uploaded to data repos where researchers can't (understandably) lose it and which have no incentive to magically make it disappear (like funders, journals, or unis).
Zenodo, OSF, figshare all have private options, so embargo isn't a valid excuse.
November 20, 2024 at 8:29 PM
Public. Data. Repositories.
Research should be uploaded to data repos where researchers can't (understandably) lose it and which have no incentive to magically make it disappear (like funders, journals, or unis).
Zenodo, OSF, figshare all have private options, so embargo isn't a valid excuse.
Research should be uploaded to data repos where researchers can't (understandably) lose it and which have no incentive to magically make it disappear (like funders, journals, or unis).
Zenodo, OSF, figshare all have private options, so embargo isn't a valid excuse.
