Keywords: The Underlying Architecture of Search

One of the main things Google uses to “see” is keywords. It works to find documents that have the specific terms you search for (or synonymous terms) in the document. Figure out the words, terms, and concepts likely to be in the document, and Google will give you what you want.

We all know what happens when you haven’t been specific about choosing keywords. You’re researching the Montgomery bus boycott that started in December 1955. You have a vague memory that before Rosa Parks’s arrest, there had been a young girl who also refused to give up her seat. But her case never achieved the iconic status that Rosa Parks’s did.

You search Google: “girl arrested in Montgomery Alabama.” Recent arrest records in Montgomery, Alabama. Sriracha, not milk.

You give it another try—it’s the Montgomery bus boycott you’re interested in? You type “Who got arrested in the Montgomery bus boycott?” Still too broad; your question lands you back at Rosa Parks.

You remember it was a teenager who got arrested. So you put the pieces together: “Who was the teenager who got arrested in the Montgomery bus boycott?” Google now zooms in on “teenager + arrested + Montgomery bus boycott,” as well as considering the connector “in” before the words “Montgomery bus boycott.” It then combs through the web looking for sources where the terms appear close together; it also compares your search sequence to that of other users and the kinds of answers that satisfied them. Aided by your specificity, you now get what you’re looking for: the name of Claudette Colvin, with links to sources like the Washington Post, the Smithsonian magazine, Stanford University’s King Institute, NPR, and the BBC.

Figure 3.16 Unspecific keywords produce unspecific results.

Figure 3.17 Keywords that are too broad.

Figure 3.18 Search results for “Who was the teenager who got arrested in the Montgomery bus boycott.”