The Movie Genome

The Movie Genome Project

Is it possible to love The Usual Suspects and hate Se7en? Glance at keywords and genres, and they look nearly identical. Both are Crime thrillers, starring Kevin Spacey, flagged for violence. Of course, one is a witty cons and scams story with a surprise twist, and the other is a disturbing, gory morality tale about a serial killer. But you've got to look inside the movies to see that.

Taste in movies is complex and individual. Yet the usual way of cataloging movies, by titles, people, and genres, flattens all this - as if you'd like a movie just because it's a Drama or stars Vince Vaughn. That's why our team of movie and TV experts created the Movie Genome, an ambitious, ongoing project with the Jinni community to map more aspects of movies, shows, and semi-professional videos than ever before - so that all different viewers can match their personal tastes and moods, and find what they really want to watch next. The Movie Genome powers the search, recommendations, Movie Personality and more on Jinni.

How It Works

Inside, the Genome is broadly divided in two: Experience - the mood and tone of the content - and Story - plot elements (One man army, Battle of the sexes), structures (Nonlinear, Story-within-a-story), flags (Violence, Nudity) and more. The Genome also includes many external aspects like awards.

The starting point of the Movie Genome is manual tagging by our team of film professionals. Each title has around fifty genes, among thousands of possibilities. Then, using advanced machine-learning technology and Natural Language Processing, Jinni's system indexes new titles automatically by analyzing user reviews and metadata. This creates a level of consistency that creative human taggers can't reach - especially important for similarity matches and recommendations, which won't work unless you compare apples to apples and battles to battles as often as possible. It also incorporates multiple perspectives (from reviews) rather than just one person's opinion. Everyone who votes on genes, as well as the Jinni team, constantly check and improve the machine tagging.

Recommendations

We think the best recommendations use man and machine. A machine can deeply analyze the type of content you like to learn about your unique taste. People can share their personal favorites and opinions about what they've seen (in a way no machine can do, as yet). Jinni isn't a social network, it's a service meant to fit how people experience media - and we've included dialogue about movies and shows as part of that.

Jinni recommends by comparing your Movie Personality and the genes of all the titles in our catalog, figuring in your preferences and some other filters. Simplistically, if you have the genes for Gloomy Love Triangles, we'll recommend titles with those dominant genes. As a user, you receive recommendations from Jinni, people you follow and your Neighbors (people Jinni identifies as having taste similar to yours).

Your Movie Personality

What can your entertainment tastes tell you about yourself? We call this Movie Personality - a complex constellation of likes and dislikes, choices and reflexes that shape your experiences of entertainment. Jinni assigns clusters of genes to each user, drawn from aspects of the content you like. These genes, along with some other measures, are continually adjusted as we learn from your ratings, reviews, and other actions on the website. We present a simplified version in your Movie Personality Sketch. No two people will have exactly the same sketch.

The Movie Personality Types offer another angle on Movie Personality, looking at characteristic attitudes and preferences that attract people to different classes of movies The Match-o-mat on any public user page compares you to other users, showing not just movies you like, but the essential traits you have in common.

Wiki - Coming Soon

Our vision of the Wiki is a constantly evolving dictionary of genes that describes numerous aspects of video, where users can shape the Movie Genome Project by suggesting and editing definitions, voting on genes, and more.