Tier 2: Full Dynamic Personalization Platform
You offer a video on-demand application for mobile devices, and want to create personalized channels
for each user based on their interests and preferences. As viewers interact
with the videos, you want the playlist to continually morph so they never run out of things to watch.
The full MediaUnbound Personalization Platform builds and maintains individual profiles for each user on a client media service, in order to help the user to explore and interact with a body of content through personalizing recommendations on the service. The system can also generate static recommendations on demand or for new users that do not yet have a profile.
Integration of Tier 2 recommendations is more complex and provides a fuller range of options for recommendation, personalization, and unique customer features. Full Personalization Platforms require dedicated hardware as these platforms are constantly updating recommendations and personalization based on user interaction with the service in question. Full Personalization Platforms are also continually updated from MediaUnbound core recommendations data. More information on our two main hosting models...
The following features are supported by this tier of the system:
The full MediaUnbound Personalization Platform builds and maintains individual profiles for each user on a client media service, in order to help the user to explore and interact with a body of content through personalizing recommendations on the service. The system can also generate static recommendations on demand or for new users that do not yet have a profile.
Integration of Tier 2 recommendations is more complex and provides a fuller range of options for recommendation, personalization, and unique customer features. Full Personalization Platforms require dedicated hardware as these platforms are constantly updating recommendations and personalization based on user interaction with the service in question. Full Personalization Platforms are also continually updated from MediaUnbound core recommendations data. More information on our two main hosting models...
The following features are supported by this tier of the system:
Personalized Recommendations
Given a user profile and a seed item, MediaUnbound can return a list of
recommended items influenced by the user’s preferences. Parameters such as
recommendation list length, item repeatability, genre filtering, and general
popularity/newness/familiarity levels can be set by the Customer. Seed inputs
can include:
- Profile only (i.e. personalized recommendations based on entire user profile)
- One or more items (i.e. artists, shows, episodes, etc.)
- One or more genres / sub-genres
- One or more of any additional catalog items
- A combination of the above
User Profiles
As a user interacts with the service (i.e. searches, plays, views, downloads,
rates, tags, etc.), raw usage data is passed to the personalization system, and
the system builds a profile of the user’s tastes and preferences in real-time.
User profiles can also be created/modified by client-side data from a user’s
local media player and library. In addition to understanding the user’s content
preferences on the item level (artist, track, video, episode, show, etc.), the
system also calculates several higher-level preferences:
- Genre preference: A score for each genre represented in a user’s interaction history.
- Popularity Index: A score describing a user’s preference for more or less popular content, generally and for each type of content item.
- Newness index: A score describing a user’s preference for newer or older items.
- Experimentalism index: A score describing the breadth of a user’s tastes and their openness to more unfamiliar, more distantly related content.
- Expertise index: A score indicating whether a user has deep knowledge of a specific type(s) of content or is a more casual fan.
 |
|
