Freesound Labs: the impact of Freesound

Freesound is a sound sharing site that was started in 2005 at the Music Technology Group. The original goal of Freesound was to collect royalty-free sounds that could be used for research and be easily shared under Creative Commons (CC) licenses with the research community. However, it quickly became a very well-known resource for all kinds of people working with audio and now is probably the biggest repository of CC-licensed sounds, with more than 300k uploaded sounds, 14k contributors and 20k downloads every day.

The impact of Freesound has surely gone beyond its use for the scientific community and way beyond our expectations. There is many people carrying out activities of very different nature that use Freesound as a resource of sounds, as a backend service or as an online community. In order to try to keep track of (at least) some of these activities that take advantage of Freesound, in April 2015 we launched Freesound Labs. Freesound Labs is a directory of projects, hacks, apps, research and other initiatives that use content from Freesound or use the Freesound API. Our main goal is to maintain this directory updated as soon as we come across new Freesound powered projects or we get notified about them, and get in this way an overview of the impact of Freesound at different levels. 

In Freesound Labs we are currently listing 28 relatively recent research papers that use and analyse data from Freesound. This data is not only limited to sounds themselves but also includes user interactions such as downloads, ratings and tagging activity. Freesound as a research resource offers a number of great opportunities as it hosts and generates large volumes of data which are sharable under CC licenses and programmatically accessible via the Freesound API. Just to name a few examples in Freesound Labs we can find research papers from fields like music information retrieval, information retrieval, natural language processing, semantic web, machine learning, auditory perception and network analysis. Freesound as a research resource also poses a number of challenges, one of the most important being how to approach data sharing and ensure reproducibility in potentially changing resources (i.e. sound descriptions and sound tags might change, their average ratings and number of downloads change over time, and sounds themselves might  even disappear if authors delete them).

Freesound Labs shows that Freesound also has an impact as a sound resource for third party applications. These applications range from music making software to games and exploration interfaces, and mainly interact with Freesound using the API (see the  apps section of Freesound Labs). Two particularly relevant examples of apps that integrate Freesound content are Ardour, a well known open-source Digital Audio Workstation, and Soundly, a tool with which sound designers can easily share sound collections. The Freesound API is therefore a great tool to increase the impact of Freesound and facilitate the reuse of its content, but it also offers a great opportunity as a tool where research outcomes can be implemented and transferred to third parties. The scalability in terms of computational resources that are required to support an intensive use of the API by different applications can however become an issue, and we need mechanisms to be able to scale in controlled and sustainable ways. 

In Freesound Labs we also list a number of educational initiatives in which Freesound is used as a backend for hosting sound recordings and for generating shared interactive sound maps out of sounds that are recorded in workshops. Some of these initiatives have originated from the Music Technology Group and Phonos, and have an impact in the society that could have not been reached otherwise if Freesound was understood as a research project only.

So far we have seen that Freesound Labs can be a useful tool to keep track of all sorts of activities and projects that happen around Freesound. However, Freesound Labs could grow and become something more ambitious than that. Freesound Labs could become a portal where creators, developers, researchers and Freesound users could share their projects and make them visible, effectively increasing the potential impact of Freesound and giving more value to the reuse of its content. Such portal would be also a source of inspiration for further works reusing content from Freesound, and would serve also as an indicator of potential new requirements for services that Freesound could research and provide.