Nanocrowd makes movie recommendations by understanding why people like movies

Movie recommendation engines are different. They deliver different suggestions because they use different methods. The two most common are:

Preference matching. Recommendations are based on comparing viewer ratings "Others who like this also liked."

Purchase history. Recommendations are based on comparing past purchases "People who bought this also bought."

Companies like Netflix and Amazon use these methods in developing their recommendations.

At Nanocrowd, our recommendations aren't based on ratings or purchases. We have developed a third approach:

Viewer reactions. Recommendations are based on people’s reactions to movies "These movies evoke similar reactions."

We analyze the words viewers use when they describe their reactions to movies. Analyzing these words allows us to understand not only what movies people enjoy, but why they enjoy them. We call this technology Reaction Mapping™.

Understanding why is crucial, because movies evoke a variety of feelings and emotions that are complex and are not captured by ratings or purchase histories. For example, when people describe Will Smith's I am Legend, they use words like "emotional, scary, mindless, zombies" and you quickly get a sense of the movie.

Nevertheless, viewers have very different reactions to I am Legend. Using our Reaction Mapping technology to create nanogenres™, we capture those differences to explain the reasons why people like this movie.

Below are three nanogenres we identified and a list of the movies we recommend in each nanogenre:


Notice how different the three lists of recommendations are. If you liked I am Legend because it's about surviving zombie attacks, then the first list is the one for you. If you like pondering the chaos of the end of mankind, then the movies in the middle list are better recommendations. You get the idea.

The movies listed for each nanogenre reflect a specific reason why people enjoy movies like I am Legend. Because the reasons are very different, so are the recommended movies.

I liked I am Legend (I've been known to enjoy a good zombie movie as much as the next guy). But most evenings I'm more of an "Intense Death Humanity" kinda guy. I was happy to see recommendations like The Green Mile and No Country for Old Men, which are perfect for me, but wouldn't show up in recommendations based on preference matching or my purchase history.

That is the reason we developed Reaction Mapping. We wanted to understand why people like movies in order to make better recommendations.

posted by Roderic March