The paper evaluates the impact of voice-controlled digital assistants on the evolution of radio services, proposing to apply radio personalisation, recommendation and content atomisation techniques to the conversational interface. The paper introduces the relevance of smart speakers for radio, and then technically describes a prototype based on voice-controlled digital assistants providing a set of personalised audio services, like keyword-based Audio-On-Demand (AoD) search, semantic similarity search and content-based audio recommendations. A first analysis of the impact of semantically coherent audio segments with metadata, or audio atoms, is also included, leveraging on a subset of the available AoD archive.

The developed prototype service is then tested and assessed with a user evaluation with the purpose to measure the listener’s reaction and propensity to access flexible radio services using a voice interface.

Introduction: Voice-controlled digital assistants and radio

The voice-controlled digital assistant market is rapidly growing and usage statistics indicate music, news and radio services among the most requested content. On the other hand, an intuitive interface based on voice and natural language radically changes the way listeners access audio content, and indirectly influence the content itself requiring different duration and formats. Currently these digital assistants are generally identified with the physical device, the smart speaker, however, the core functionality they provide is AI. After this early wave, digital assistants will probably spread widely in a variety of different devices, see the report by Activate.

Recent market analysis estimates show that awareness and ownership of smart speakers is rapidly increasing, with respectively 223 million people being aware of the technology and 65 million people owning one device or more only in the USA in 2019, see Edison Research and Triton Digital and NPR and Edison Research. Currently the biggest players are Amazon and Google, see Figure 1, but from 2014, the year of introduction of the first device from Amazon, several other companies have brought their voice-controlled digital assistants to the market: Apple, Samsung, Huawei, Baidu, Xiaomi, Alibaba and others (all product names and brands are property of their respective owners). The devices are showing the fastest adoption rate of any consumer device, and the functionalities they provide will probably stay with us for a long time, see (1). On the other hand, using smart speakers, listeners often ask for music (90% of first adopters in the last week), news (51%), live radio stations (41%), see (3). Also, a part of smart speaker owners uses them instead of traditional radios (about 40%).

The above scenario makes smart speakers extremely relevant for radio. This fact and the awareness that a third party who owns the digital assistant can act as a gatekeeper, make them both an opportunity and a threat for radio broadcasters. Analysing the potential impact of these devices is therefore useful for radio service evolution. Several European broadcasters have proposed prototypes and services for radio on smart speakers in recent months, and the EBU has created the VOX expert group to foster collaboration.

As a step towards the understanding of how to leverage the power of smart speakers in favour of radio, we designed a prototype, working with content from Radio Rai. The selected platform was Google Home, as it was the first smart speaker launched in Italy in March 2018.

Download the full paper below