Beautiful Visualization

Chapter 9. The Big Picture: Search and Discovery

Todd Holloway

SEARCH AND DISCOVERY ARE TWO STYLES OF INFORMATION RETRIEVAL. Search is a familiar modality, well exemplified by Google and other web search engines. While there is a discovery aspect to search engines, there are more straightforward examples of discovery systems, such as product recommendations on Amazon and movie recommendations on Netflix.

These two types of retrieval systems have in common that they can be incredibly complex under the hood. The results they provide may depend not only on the content of the query and the items being retrieved, but also on the collective behavior of the system's users. For example, how and what movies you rate on Netflix will influence what movies are recommended to other users, and on Amazon, reviewing a book, buying a book, or even adding a book to your cart but later removing it can affect the recommendations given to others. Similarly, with Google, when you click on a result—or, for that matter, don't click on a result—that behavior impacts future search results.

One consequence of this complexity is difficulty in explaining system behavior. We primarily rely on performance metrics to quantify the success or failure of retrieval results, or to tell us which variations of a system work better than others. Such metrics allow the system to be continuously improved upon.

A supplementary approach to understanding the behavior of these systems is to use information visualization. With visualization, ...