Jinni

Jinni’s discovery-and-recommendation engine for premium video content integrates semantic search, browse and personalized recommendations, satisfying the widest range of customer requirements and user mindsets. Based on content genetics and user taste profiling, Jinni’s discovery engine powers a uniquely intuitive, personalized experience via a plugin to thePlatform’s Player Development Kit (PDK).

Jinni’s intelligent discovery service is powered by its semantic catalog – the “Movie Genome” – which was designed specifically to power content discovery. (Unlike traditional genres, which were designed for back-office cataloging.) The taxonomy was developed by film professionals. New titles are indexed automatically using innovative Natural Language Processing technology to analyze user reviews, synopses and metadata. Jinni assigns semantic tags to both content and users, and utilizes clustering, similarity and relevancy analysis algorithms to power discovery and personalization.

Using its proprietary semantic tags, Jinni creates a one-of-a-kind taste profile, or “Movie Personality,” for each user. The profile is based on the user’s ratings and preferences, and can also incorporate implicit data. This approach is notably different from collaborative filtering, which uses statistical correlations between users and content items, and cannot easily present the “whole picture” of a user’s taste. Jinni user taste profiles can be explained in natural language, adjusted based on the user’s ongoing interactions, compared to other users’, and used as a tool to facilitate content discovery. CF is employed as a complementary tool.

The Jinni Web Services API is a modular content discovery solution that makes accessible all the features of Jinni’s discovery-and-recommendation engine, for customers delivering content across multiple platforms, to STBs, PCs, and mobile devices. The API is designed to integrate into the systems of Jinni’s customers and partners, and to work on their catalogs and other data.

Jinni provides a complete discovery-and-recommendation engine for TV operators delivering content to STBs, PCs, and mobile phones via a comprehensive set of APIs. The service has been designed to enable the operator to develop a next-generation guide with a superior user experience and taste profiling, to increase consumption and reduce churn while addressing commercial objectives, branding and solution.

Recommendations done the Jinni way are accurate (the user is likely to enjoy watching the recommended titles); effective, coverage-wise (all titles in the catalog are considered); diverse (representing different topics, styles, moods etc. that are within the user’s tastes); interesting (obvious recommendations are avoided and surprising ones are occasionally given – the so-called exploration/serendipity effect); and explainable (the algorithm choices can be justified in natural terms, using a rich set of Genome-based expressions).

A taste-based Movie Personality also supports community functions. The Jinni system can identify “neighbors” for a user, explain the compatibility using semantic Genome terms and suggest content items both users are likely to enjoy.