Monday, May 21, 2012

Google Knowledge Graph, Taxonomies, and you.

Google has just released a new search experience for their end users called the Google Knowledge Graph. Knowledge Graph takes advantage of relationships between entities that it has deduced (either explicitly with a taxonomy or benefiting from its large volume of search history and click behavior) to present more relevant search results to users. So, a search for Leonardo Da Vinci will return a right hand panel with links to works by Da Vinci, birth/death date, Wikipedia description etc. The idea is to get you to an answer more quickly. A search for the Denver Broncos will surface information about the roster, head coach, mascot, etc.

In my initial surface testing, it works well for proper nouns (city names, people, teams, movies, books) but for searches like "Taxonomy" there are no Knowledge Graph results, just the normal search listings. I'm sure this will get more robust over time

The main value seems to be that Google is pulling relevant information about a topic that somebody has searched for and making it explicit to the end user as well as providing some additional search result filters or other suggested searches.

This is relevant to enterprise search because Google is setting the bar for how people expect their search experience to operate. I've spoken to many information architects who are trying to provide their users with an easy to use Google style search, because that is what the users are asking for. What users don't often realize is that a google style search doesn't necessarily work very well inside the enterprise; the concept of using links to establish relevance of a website doesn't work to establish relevance of a document because documents are not linked together. However, the concept of tagging documents with relevant business metadata and surfacing those tags as refinement and browsing options to enhance a keyword style search works very well inside of an enterprise. This is where taxonomy comes in.

Google Knowledge Graph is going to train users to expect the type of enterprise search experience that only a good, relevant taxonomy and tagging strategy can make possible. Whereas Google Knowledge Graph is using a robust ontology of all topics to surface such information, an enterprise just needs a good taxonomy (or taxonomies) of topics important to their business. It's a very achievable task (particularly if you have a good taxonomy to start with).

Google has just helped educate the masses - taxonomy is about to get big.