Siri & HomePod
Siri & HomePod
A voice first device.
A voice first device.

COMPANY: APPLE

COMPANY: APPLE

COMPANY: APPLE

CATEGORY: ARTIFICIAL INTELLIGENCE

CATEGORY: ARTIFICIAL INTELLIGENCE

CATEGORY: ARTIFICIAL INTELLIGENCE

HomePod
HomePod
HomePod

Harry Simmonds, the founder of Studio Elros, played an integral role in shipping Apple's voice-first product, HomePod. As a premium music device, HomePod required Siri to develop an advanced knowledge of music and interact with users to access millions of songs. Intelligent assistants represent the pinnacle of artificial intelligence, and using them as the primary interface of a product presents an incredible challenge.

What was done

  • Personalised music integration 

  • Enhanced music library search

  • Synchronised music across the home [patent granted] 

  • Increase reliability and accuracy of music search queries by voice 

  • User corrections of incorrect or undesired music

01.

Focus on core user experiences

Voice interfaces are inherently imprecise, and we tend to interact with them in much the same way as we do in our everyday conversations. However, artificial intelligence has not yet reached the level of grammatical and semantic understanding that an average person possesses. So, how can we focus on the core experience and educate users accordingly, especially given the open-ended nature of voice interaction?


To address this challenge, we combined our user studies, user research, and usage data analysis to focus on frequently asked queries and domains. Through this analysis, we identified several user features, such as personalised music and home-wide synchronisation with HomePod.

Personalised Music

Before the music industry transitioned to a streaming model, we used to spend countless hours downloading our favourite songs, curating our own playlists, and developing a close relationship with the music we owned. Streaming has transformed this experience, allowing for new and different ways of enjoying music, such as recommended playlists and stations. Voice interfaces are particularly well-suited for music recommendations. You can simply request the type of music or artist you want to listen to, and the system will generate similar music, making it much easier to discover new music.


For the release of HomePod, Harry focused on designing a system and architecture that would retrieve personalised music from Apple Music through voice requests, creating a novel way of listening to music in the home.

Apple Music Top Picks
Apple Music Top Picks

Home Wide Synchronisation

Put a HomePod in your living room and bedroom and you start to have music for the home. However, using voice commands to control home-wide music presents a unique challenge for HomePod. The language used is often generic, and as we discovered through our user studies, users expect Siri to understand these references. For example, when users say "Play it here" or "Play it everywhere," it's not always clear what "it" refers to and what "here" means. To address this challenge, Harry spearheaded the software design to associate music with a room and a room with a specific verbal reference. This allowed users to request synchronising and differentiating music across their home.

02.

The necessity of fault tolerance for a better user experience

Although the accuracy of voice interfaces is not perfect, their powerful applications make them highly valuable. However, when a misrecognition or misunderstanding occurs, how do we ease the user's frustration? Additionally, how can we provide alternate choices to the user?

Alternative Corrections

With over 40 million songs in its library, Apple Music likely has multiple versions of the same song, such as "Shake It Off." While search prioritisation can help alleviate most issues, there are situations where we want to ensure that the user can easily correct to their desired music content. For example, the user may want to listen to the acoustic version or discover niche music that doesn't always surface from popular hits. Through our research, we developed voice corrections that allow the user to request, "No, I want the acoustic version" or "No, the other version," enabling a more seamless and personalised music experience.

03.

Improvement via data analysis and metrics 

In our industry, it's common to focus on core experiences when designing a product. However, for a product with infinite input possibilities, it's crucial to carefully sort through all the data and find the most valuable insights. But how do we label these immensely complex data sets? For example, how can we determine if an error was caused by voice recognition, semantic understanding, the music search engine, user error, or personal music preferences?


These are the problems our teams have focused on, working across the entire Siri architecture to classify errors and gain insights into areas for improvement. By identifying key performance indicators (KPIs), we were able to prioritise improvements that would have the most significant impact on user experience. Through this process, we have gained valuable knowledge about the performance of our product, enabling us to continually refine and improve the Siri experience.

Media Credit: Apple Inc.

Let’s talk

Subscribe to our upcoming newsletter for occasional updates! We only send emails when we have something to say.

© 2023 Studio Elros

·

Let’s talk

Subscribe to our upcoming newsletter for occasional updates! We only send emails when we have something to say.

© 2023 Studio Elros

·

Let’s talk

Subscribe to our upcoming newsletter for occasional updates! We only send emails when we have something to say.

© 2023 Studio Elros

·