For some, the idea of computers being able to comprehend human levels of understanding is a fear-inducing, apocalyptical vision. For others, it’s simply logical progress. Well, depending on your point of view, Google either has great news or very bad news – the Knowledge Graph is here.
What’s the Knowledge Graph?
Well, some are calling it the ‘dawning of the semantic web’. Let’s knock that on the head straight away, it’s not. However, it is a way of the search engine attaching meaning to words beyond a single definition and finding connections within a query. What?
There are some words and phrases that are pretty straightforward when it comes to providing search results. Let’s take someone who’s in the news currently as a good example, Kenny Dalglish.
As Liverpool fans will testify, there is only one Kenny Dalglish (of note), therefore when somebody searches for his name, there’s a good chance that they want to find news and information related to this single person. But Kenny Dalglish is not a one dimensional character, he is connected to many events, individuals, dates and so on.
Therefore, Google will use its Knowledge Graph to connect the dots. By plundering data from a whole host of sources, including Wikipedia and Freebase, they will be able to find relationships with a whole host of other data. For instance, Kenny Dalglish was the sole representative from Scotland on the FIFA 100 list. This in itself provides multiple connections.
Firstly he’s Scottish and represented the national football team. He’s also connected with the 99 other footballers on the list. Of course, using the six degrees of separation rule, he might well have associations with millions of people, events, places and the like. However, if you were to search for “who was included in the FIFA 100” he should appear, with related information, in search results.
This form of computational connection is nothing new though. Users of Wolfram Alpha have been able to access and provide answers using connected information for a while now, albeit in a slightly more limited capacity. Essentially, the ultimate challenge is to be able to use the data available to provide answers to complicated questions, not just piece together keywords.
When you use keywords to determine relevance, it’s easy to lose meaning. Let’s take a generic word with multiple potential meanings, like “Polo”. Without other terms to provide context, it could relate to any of the following
Polo – the equine-based sport
Polo – the mint
Polo – the Volkswagen
Polo – shirt
At the moment, you get results that look a little muddled:
By categorising these individually and understanding the context of connected words such as “Events”, “Ascot Park”, “1.2”, “Ralph Lauren”, Google can begin piecing together what you mean. However, the new Google interface will now provide an option to search within the overarching phrase “Polo”. So, using the “See Results About” which will feature on the right-hand side of the screen (just like maps generally do), you can filter the results and categorise them appropriately.
Whilst not entirely groundbreaking, this demonstrates that the search engine isn’t seeing the word as a singular term, but is aware of alternative possibilities. This requires knowledge and a deeper understanding of context.
At the moment you can source sports results, flight information and weather updates through a search. These are pretty basic searches though, that rely on short, generic queries and a single source of information. That’s not to say that it isn’t useful, but the Knowledge Graph will go above and beyond this, dragging in information to answer questions with a human-level of understanding.
So how would this work? Well, let’s imagine you ask “When did Portsmouth reach their first FA Cup final?” It’s a pretty basic question for a human to answer, even if you don’t have the information. Most would simply look for “Portsmouth FC” of “FA Cup” online and visit one of the many websites that has detailed information about their history. From this you should be able to gauge when they first got to an FA Cup final. The answer is the 1928-1929 season, where they lost 2-0 to Bolton Wanderers.
As search currently works, Google would use the keywords “FA Cup”, “Final”, “Portsmouth” to provide sites relating to those terms. In future though, it may be possible for them to use the Knowledge Graph to connect these three and provide a history of Portsmouth reaching FA Cup finals, perhaps even working out the chronological order of these. As the data is all available on Wikipedia, it simply takes an understanding of the query to match everything together.
It’s all very complicated stuff, but it could well be a huge step forward in how search engines understand language and connections. Like a human brain, Google may take some time to build up this knowledge and become more efficient and accurate with responses. It also appears that they are going to become even more reliant on Wikipedia as a key source, potentially upsetting other folk who have been receiving decent traffic for related search queries in the past.
For research purposes, the ability to really drill down into a subject, including a range of media, sources and information, from a single source could prove to be hugely beneficial. Whilst there are certain semantic elements to this, it’s not a complete overhaul of search by any means. Google claim that the Knowledge Graph will feature on as many results pages as maps are currently shown, which is a fair percentage, but is by no means across the board.
For this reason, it can’t be viewed as being the perfect example of the semantic web in operation. Perhaps it is a major stride in the right direction, particularly when you consider this is the first time Google has attempted to go beyond keyword phrases, but it’s not the end of the road just yet.
Nevertheless, this is a major development and one which we are likely to look back on in years to come. It is also another nail in the coffin of keywords, even if most are still delivering strong traffic levels. Google are attempting to navigate away from this style of indexing sites, either through semantics, personalisation or social influence.
For more information about the update and how it will impact SERPs visit the Google blog here.