Exploring Internet Radio Across the Globe with the MIRAGE Online Dashboard

2025

Read Exploring Internet Radio Across the Globe with the MIRAGE Online Dashboard (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2502-05250.txt.

Introduction

Ngan Nguyen, Elizabeth Acosta, Tommy Dang, and David Sears present a 2025 paper on a web dashboard for exploring internet radio across the world. Radio streams carry rich records of what music plays where. Those records are hard to gather and search at a global scale.

The dashboard, called MIRAGE, covers one million streaming events on ten thousand stations. Each event logs the artist and track, plus music features such as instruments, voice type, and key. Users can search by station or event, view charts that respond to clicks, and listen through links to streaming services. They can also export the data and charts for their own study.

Catalog record

YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE25 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-05250

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Exploring Internet Radio Across the Globe with the MIRAGE Online Dashboard Abstract This study presents the Music Informatics for Radio Across the GlobE (MIRAGE) online dashboard, which allows users to access, interact with, and export metadata (e.g., artist name, track title) and musicological features (e.g., instrument list, voice type, key/mode) for 1 million events streaming on 10,000 internet radio stations across the globe. Users can search for stations or events according to several criteria, display, analyze, and listen to the selected station/event lists using interactive visualizations that include embedded links to streaming services, and finally export relevant metadata and visualizations for further study. 1 Introduction Despite its scholarly neglect relative to television, film, and print [ 1 ] , radio’s convergence with the internet has extended its reach via web browsers and smartphone apps, enabling the medium to persist as a central site of culture and daily life for communities around the world [ 2 , 3 ] . The recent resurgence of pirate and community radio stations on the internet alongside national and multinational networks also reflects internet radio’s lower production costs relative to short-wave terrestrial (e.g., FM or AM) radio [ 4 ] , resulting in a diverse range of both standardized and specialized programming [ 5 , 6 , 7 ] . And yet, the volume and scope of much of the research in fields like radio studies has been freighted heavily towards the Global North [ 1 ] . In doing so, the research program just described attempts to situate listeners within a particular musical tradition (e.g., western classical or popular music), rather than within a particular geographic environment (e.g., El Paso, Texas) where myriad musical traditions might co-exist. As a result, music’s vast global marketplace has yet to receive sustained scholarly attention in the MIR community. To address this issue, this study presents the development release (v0.2) of the Music Informatics for Radio Across the GlobE (MIRAGE) online dashboard, which allows users with potentially little training in computational methods to access, interact with, and export metadata (e.g., artist name, track title) and musicological features (e.g., instrument list, voice type, key/mode) for 1 million events streaming on 10,000 internet radio stations across the globe. To that end, Section 2 summarizes previous research on the development of digitized music corpora and cultural databases. Next, Section 3 presents the MIRAGE-MetaCorpus, Section 4 introduces the MIRAGE online dashboard, and Section 5 offers a potential use case. Finally, Section 6 discusses limitations and future directions for the MIRAGE project. 2 Previous Research In recent years, researchers in music theory, music information retrieval (MIR), and radio/media studies have developed digitized music corpora and cultural databases that represent data in machine-readable symbolic and audio formats. Radio Sovereign Continent Stations States Africa 392 (4%) 39 (58%) Asia 653 (7%) 38 (59%) Europe 5,161 (52%) 49 (60%) North America 2,243 (22%) 33 (67%) Oceania 222 (2%) 5 (17%) South America 1,329 (13%) 13 (87%) TOTAL 10,000 177 Figure 1 : Descriptive statistics (left) and geographic map (right) of the radio stations in MIRAGE-MetaCorpus. The size of each bubble represents the number of stations at that location. In computational music theory, heavily curated corpora (100s of songs) like the McGill Billboard and Rolling Stone-200 data sets include expert annotations for musical parameters like harmony, meter, and melody, for example, but remain restricted to Anglophone popular music traditions [ 8 , 9 ] . What is more, the limited size of symbolic corpora makes comparative research especially difficult [ 10 ] . In MIR, corpora like the Million Song data set address issues of scale while avoiding copyright infringement by providing researchers with publicly available metadata and musicological features protected under fair use for a large collection of songs hosted on commercial music-streaming services like last.fm [ 11 ] . Nevertheless, the size, scope, and format of these projects require extensive training in distant-reading (i.e., computational) methods [ 12 , 13 , 14 ] . As a result, MIR corpora sometimes eschew the kinds of musical engagements favored by scholars in humanities disciplines using close-reading methodologies. Finally, the projects referenced above do not include information about the geographic location of the music encountered by listeners in everyday life. Finally, in radio/media studies, researchers routinely employ interview and survey methodologies to explore radio stations across the globe [ 15 , 16 ] , in some cases by selecting samples from radio-station directories hosted online [ 17 ] . The now defunct ComFM, for example, included a catalogue of web-radio stations classified according to geographic region and type of programming. Other current internet radio directories like radio.co and internet-radio.com offer searchable databases consisting of several thousand stations, but they do not permit users to access or export the entire database for further analysis. The MIRAGE online dashboard addresses these issues by offering a global archive of the musical traditions encountered on internet radio. For this reason, the dashboard’s database could serve MIR tasks like music recommendation and genre classification, but the dashboard itself also allows researchers with potentially little training in computational methods to select and analyze a subset of events or stations (i.e., to develop their own sub-corpora). Finally, like previous MIR projects [ 11 ] , the MIRAGE online dashboard avoids copyright infringement by including publicly available metadata and musicological features protected under fair use while enabling users to stream recordings using embedded links to commercial services like Spotify and YouTube. 3 MIRAGE MetaCorpus The core database for the MIRAGE online dashboard is MIRAGE-MetaCorpus, which currently consists of metadata and musicological features for 1 million events that streamed on 10,000 internet radio stations across the globe. In this context, an ‘event’ could represent a musical work of some kind, or a radio program like a podcast or a call-in show. 3.1 Collecting MetaData Following [ 12 ] , data collection consisted of three stages: station-list and event-list collection ( Stage 1 ), station-list review ( Stage 2 ), and event-list parsing ( Stage 3 ). 3.1.1 Stage 1: Collecting Station/Event Lists Toward Stage 1, the research team collected metadata for an initial list of internet radio stations and then monitored the station streams to obtain additional metadata from the stream encoder. To that end, we monitored radio stations in real time on Radio Garden, 1 1 1 https://radio.garden . a streaming service with an open-access application programming interface (API) that allows users to select and play publicly available radio streams using an interactive representation of the globe. Between the months October to January 2022-2023, a random sample of 10,000 stations from the initial station list was monitored throughout the 24-hour day – but avoiding each ten-minute period at the top and bottom of the hour when advertising is most frequent – in order to obtain additional metadata from the stream encoder for 100 events from each station, resulting in an initial list of 1 million events. The monitoring algorithm also excluded an event if the stream description did not include metadata, or if the metadata featured advertising terms or reflected a station blackout period (e.g., ‘advert’, ‘commercial’, ‘unknown’, ‘blackout’, etc.). During event-list collection, additional metadata for each location in the initial station list was also included

Read the complete text