Music Discovery Engine

Our approach to Decentralized Music Discovery presents a significant innovation over existing systems like the Music Genome Project, which is still in use by millions of users of Pandora, a SiriusXM-owned company. The Music Genome Project relies on a vector-based system, where songs are analyzed and assigned various attributes (or "genes"), such as vocal timbre, instrumentation, and mood. Recommendations are then based on matching these predefined attributes between new songs and those from users' listening histories. This method is relatively static (as the predefined attributes do not change) and does not evolve dynamically based on real-time user behavior.

In contrast, MixerFM's Decentralized Music Discovery system is powered by a graph-based architecture that adapts to user preferences dynamically. In this system, users and music mixes are represented as nodes in a graph database, and their interactions are connected by edges that carry a dynamic weight.

How it Works:

• Nodes and Groups: Users (e.g., Ana and Jeffrey) and music mixes (e.g., Mix ABC, Mix DEF) are represented as nodes in the system. The system groups them based on shared preferences or interactions, which helps MixerFM understand commonalities in user tastes. This flexible grouping is more dynamic than the fixed vectors used in traditional systems.

• Edge and Weight: The connections (edges) between nodes carry a weight that reflects how much a user interacts with or likes a particular piece of content. This weight can range from -1 (strong dislike) to 1 (high preference), with 0 representing neutrality. Positive interactions increase the weight, making it more likely for that content to be recommended again. In contrast, the Music Genome Project's system doesn't allow for real-time adjustments based on user interaction.

• Interaction-Driven Weighting: Each user interaction, whether liking, commenting, or skipping a track, modifies the weight of the connection between the user and the content. For example, if Ana frequently engages positively with "Mix ABC," the system increases the weight of that connection, making it more likely for Ana to receive similar recommendations. This real-time adaptability significantly enhances personalization.

• Chain of Context: The Chain of Context is the backbone of MixerFM’s recommendation engine. It captures and records all user interactions: likes, comments, tags, and shares to create a context-aware system that adapts recommendations based on multi-layered inputs. Unlike traditional platforms where these interactions are hidden from the user, the Chain of Context ensures that every action is tracked transparently, giving users control over how their behavior influences recommendations. As users interact with content, leave comments or tags, and create AI-assisted mixes, these inputs directly shape the context in which content is recommended. The Chain of Context allows for close to real-time adjustments.

The MixerFM Recommendation Engine evolves from traditional music discovery systems like Spotify or Apple Music, which often rely on static, popularity-driven algorithms. At its core, MixerFM dynamically constructs multidimensional user profiles by utilizing near real-time interactions, emotional context, and user inputs. This approach ensures that recommendations adapt to social cues and emotional states in near real time.

Verifiable Song Ratings and Sentiment-Based Feedback

In contrast to systems that use basic like/dislike functions, MixerFM incorporates sentiment-based signals, gathering real-time data from users’ interactions, such as context-rich searches and social posts. For example, if a user searches for mixes using the query “light-night heartbreak” or expresses feelings like "I feel like a main character" in a social comment on the last track/mix they streamed the engine dynamically adjusts recommendations to match those emotional tones.

Multi-Platform Integration TBD

MixerFM plans to offer multi-platform integration as a future expansion, allowing other decentralized apps (dApps) to access its recommendation engine and user data powered by its decentralized infrastructure. This will enable dApps across different platforms to sync and leverage MixerFM’s data, creating a more connected ecosystem. This expansion will allow other dApps to offer contextually aware, personalized recommendations to their users, helping MixerFM become a core music discovery service in the decentralized space.

Key Differences in Our Approach

More Than Just Likes:

MixerFM integrates all user inputs, from comments to emoticons, to give a complete picture of user preferences.

Personalization:

Recommendations are adjusted in near real-time based on multi-level feedback, ensuring they stay relevant.

Emotionally-Aware Recommendations:

By capturing a range of emotions through emoticons and comments, MixerFM delivers content that resonates on a deeper level.

Transparent Recommendation Process:

The Chain of Context ensures users can see how their interactions influence what they are recommended, building trust through transparency.