A couple of days ago, Gianluca Fiorelli published a thoughtful look at the Search Industry in past year, and the year to come at Moz titled SEO and Digital Trends in 2017. He included a graphic within that which listed things that he considered important events in the industry, including patents that had been granted in 2016. He listed patents that I had written about in that graphic, but hadn’t linked to them in the post, so I considered doing so, and mentioned in the comments that I likely would. I also wrote a number of posts on the Go Fish Digital Blog, and decided that I would link to some of the ones that were granted in 2016 as well.
Here are the 2016 patents granted to Google that I thought were interesting enough to write about this year, and something about what they do:
Google describes how they might identify and map search suggestions for different entities, by looking at properties associated with those entities.
Google describes how they may use News sites as sources for featured snippets, and use algorithms that can create shortened paraphrases to generate answers for those featured snippets.
The patent that Google came out with and described how PageRank might be modified by looking at the features and characteristics of links on pages, to understand what seemed to be the most important links was rewritten with a little more emphasis on how important the anchor text in those links appeared to be, and how likely it may be that someone would click upon that link.
This points to the possibility that someone might see an entity on a webpage, be be given an option to interact with that entity, such as being able to make a reservation with a restaurant, or being able to get driving directions to a business.
A patent that describes how Google used machine learning to better perform customer service for their Adwords system.
Ranking events in search results could be based upon PageRank and Information Retrieval Scores, but those events could also be ranked based upon how popular entities holding those events might be, and how popular the venues are were they are held. I like ranking based upon the entities and venues more, myself.
A horse might mean an animal to an equestrian, a tool to a carpenter, or an exercise implement to a gymnast; if a search engine counted up the mentions of a horse in a knowledge base under each type of meaning, it could use those numbers to create vectors that could help it index content based upon multiple meanings of words, and use that knowledge to better understand the meanings of queries; resulting in a smarter search system.
Google might identify topics and properties of entities from knowledge bases, and use that information to better verify facts on web pages when possibly recommending those pages as “news”.
Local search might start using a distance from your location history than a distance from your desktop computer as a way to rank places that you might find in Google Maps.
A Google patent describe changes to the Keyword Suggestion Planner tool from Google.
A patent from Google describes how Google Glass could be used to recognize songs and display the lyrics to those songs.
Google might try to understand the different entities that appear in a set of search results and cluster those together.
Rankings of entities that appear in search results could be done based upon the sentiment associated with those entities.
We are given a view of some possible additions to Google Maps in this patent, including: Carpool matching, Location-based reminders, Recommendations for businesses, Special events triggering alerts.
Imagine walking up to a restaurant that you don’t know how to pronounce the name of (on a vacation to another country, maybe), and you search using a phone for the word “reviews” to see reviews about it, and the search engine guesses that you mean the restaurant right in front of you.
Businesses that have a local significance to them may be boosted in search results, and places that have a global significance may be demoted in search results.
A patent that describes how an entity identification model might work at Google to help it better recognize entities.
Some queries are specifically asking for entities (who directed Star Wars?) while other queries are asking for descriptions (what is Star Wars about?). If Google is able to better classify whether a query is asking for an entity or a description, it can tell if it has given a good answer better.