Back
R&D
Christmas 2016 had consecrated the Amazon Echo kiosk as the most popular gift in the United States, 2017 saw the arrival of Google Home, and signs the integration of Google Assistant in the TV. A recent study by Strategy Analytics speaks of 24 million smart speakers that will be shipped worldwide in 2017. Amazon and Google share a market that will soon diversify: Facebook's intelligent assistant, "M" is for the moment still limited to Messenger, but the firm is planning to launch an independent device. The Apple HomePod focuses mainly on music, but the brand is making its SiriKit available to all iOS developers. Logitech launches the first connected intelligent nomadic loudspeaker, while Invoke, Bixby, Sonos, Lenovo also want to have a voice in the chapter.
In France, the market is still relatively small: only Google Home is available and the first French applications compatible with Google Home and Google Assistant have been released since the end of October 2017. Amazon (71% market share across the Atlantic) plans to launch in the language of Molière in the 1st quarter of 2018. Orange, in partnership with Deutsche Telekom, is trying to break the hegemony of the Gafa with its assistant Djingo, also scheduled for release in early 2018.
Understand each other better
The rise of voice assistants is mainly due to the improvement of speech recognition. Whereas the success rate for voice interpretation was less than 80% in 2013, Google's voice recognition has been within 5% margin of error since last year thanks to advances in NLP (Natural Language Processing). When a 99% success rate is achieved, the market is likely to shift to an even higher speed.
Today, 20% of mobile searches are made by voice. By 2020, according to Comscore, it will increase to 50%. Even if some of the assistants' voices may remind us of the old Star Trek robot, the technology is already capable of imitating human voices, and may soon create links as passionate as that of Joaquin Phoenix with his digital assistant Samantha in the film "Her" in 2013.
What is the difference with the classic keyword search?
The first use seems obvious: searching by voice mostly takes over when our hands are busy, for example while driving. But already in 2016, Mary Meeker cited a report in which, while 36% of respondents used voice commands in the car, 43% said their first use was at home. We all know how difficult it is to use a remote control when we want to go beyond the classic zapping between channels: voice seems to be a privileged companion in this type of use.
Why use a voice interface?
Voice research also has the advantage of being more precise: because it is conversational, it lasts longer, and therefore reveals more details about the researcher's intention (and therefore more information about his personality to propose adapted content).
What place for a media?
Of course, personalization through the mastery of algorithms and big data already allows us to offer the public the product that best suits them, thanks to a certain number of criteria such as their behavior, their appetites, their friends' tastes, the most viewed programs, or the most sponsored. But voice piloting makes content discovery even more accessible.
An analysis of the Hound app differentiated 4 categories of research, including one particularly interesting for France Télévisions : Personal Assistant (27% of app usage), Entertainment (21% with access to videos, including purchases), General Information (22%) and Local News (22%).
LIA: France Télévisions tests its own voice assistant
To test the fields of application of this new interface, France Télévisions has partnered in early 2017 with the start-up LLC One to experiment a proprietary voice assistant, code name: LIA.
The project includes two phases:
Phase 1: How can we interact with France Télévisions services: launch a video, search for a news article, find a program?
Phase 2: How can we interface with products on the market (Amazon, Google, ...)?
Interaction with France Télévisions content
The LIA conversational service has been developed on an iOS environment. It allows you to launch video services by voice (live, replay, future SVOD catalog) and to carry out searches with word-precision (a name, a brand, etc.). It also gives the possibility to search for an extract among all available replays.
Demonstration of the use in the video below ⬇️
The stakes for France Télévisions are multiple:
The voice assistant will make it possible to provide a new service to the public. It transforms the perspectives of video consumption, already largely delinearized. It will also provide an answer in terms of accessibility.
Feedback from phase 1 of the test: the reactions of internal and external testers (presentation at the RG Lab at Roland Garros) are positive, the service is fast, accurate, relevant and intuitive.
Experimentation is still ongoing, with the ultimate goal of transforming LIA into a true conversational service beyond simple voice control. A service that dialogues with the viewer, capable of qualifying queries and offering several possible results, all in a more user-friendly way than a simple search bar, a bit like replacing the speakerines that have long since disappeared from our screens.
Interfaced with France Télévisions' DMP and CRM, it should enable relevant recommendations and better loyalty. The objective: to make LIA the personalized guide to France Télévisions' content.
Another improvement in progress: the basic synthetic voice will be replaced by a more advanced voice.
Interfacing with products on the market
Phase 2 of the test will consist of moving to a multi-platform service (mobile, desktop + multi OS), available on the largest number of devices on the market. LIA will have to open its APIs to Amazon, Google, etc. so that the assistants in the connected mobiles and speakers can launch videos and play them on their TVs without any further interaction. Tests with some platforms are underway.
For France Télévisions, it is important to master the technology so as not to be locked into the turnkey solutions proposed by the market giants. Mastery of technology allows us to be more at the service of our audience, to offer personalized solutions and to innovate.
Conclusion
Voice is a major issue for FTV: this new use must be put at the heart of the products, and will multiply the offer by making it possible to navigate from one product to another. LIA will evolve to be both present in our services and connected to the new uses that emerge every day, for example the autonomous car. Beyond searching by voice, audiovisual media must also find a way to exploit the full richness of their image catalog. What new uses should we invent, and how can we make the most of our content on these voice-centric platforms, while waiting for the arrival of connected speakers with screens?
Voice assistants are an additional platform on which the media will have to be present, innovating and optimizing content production processes. But they are also a great opportunity to create a link with the viewer, to enhance the value of programs by making life easier for them while giving them a choice. To get started, it is essential to maintain a methodological and strategic approach that requires putting the end user and his or her protection at the heart of the reflection. As with every new technology, we must start from the story we want to tell, not from the tool. It took us five to ten years to learn the subtleties of dialogue with our touch screens, it won't take much more for the voice?
