Find Results Faster with Visual Search

3 min readFeb 23, 2016

Search engines and recommendation engines display results ranked by relevance, in a list or a thumbnail layout. That’s useful in many cases but sometimes it’s more useful to also explore the relation between results. Consider the human mind. When we try to remember something we hop between bits of memories, one memory leads to another, and using associative thinking we eventually find what we are looking for.

Here’s a joke about associative thinking. Two couples meet for dinner and while the women leave for the living room the men stay at dinning table and talk. “Hey” says one “I ate at a great restaurant the other day. What was the name? Something to do with flowers. Daisy, Blossom. No I got it.” Then he shouts to the living room: “Rose, what was the name of that restaurant?”

For that propose Natali, Roi and I built Kluster a search tool for associative exploration. The name comes from a combination of cluster and K-means, a popular clustering algorithm that we used in an earlier version of the tool. With Kluster you can choose one of several paths of exploration, and based on the path you choose different results will reveal themselves.

Let’s say that you like the book Zen and the Art of Motorcycle Maintenance and you look for similar books in a regular book recommendation engine. It will show you other books about Zen and about motorcycles. It will also show you books about workshops and maintenance, philosophy, journeys and other classics from the seventies. But why did YOU like the book? I for one, don’t care much about zen and certainly not about motorcycles. I am interested in philosophy and journey books. You might have another preference. Whatever your taste, Kluster separates results into automatic categories, allowing you to focus on what interests you the most.

1. Here I’m starting a search with Kluster with Zen and the Art of Motorcycle Maintenance, and I get a few recommendations: