Omnity Semantic Search Engine

It’s true that when you’re at the top everyone wants to pull you down. This is the same for brands as it is for people and Google is no exception. Part of this is natural selection, you automatically become the ‘target’ to beat. As a trailblazer your achievements become the benchmark others measure themselves by. 

While this may seem a tad deplorable it is also useful. Top slots are always finite by nature which means that there will always be contenders willing to try for them. This keeps the guys at the top sharp and helps them maintain their skills and it makes the guys coming up from the bottom try that much harder because getting to the top slot is no longer easy. 

Google has ruled search for a very long time and search has become the ubiquitous technology we use to successfully navigate an increasingly large sea of data: 

https://www.youtube.com/watch?v=O9hQ3D3wWFM

When so many people rely on search it also follows that search can become the key to great wealth. This attracts a lot of attention. Despite the fact that the public failure of past industry giants like Microsoft and Yahoo! to unseat Google showed just how hard it is to do search well, there has never been a lack of contenders. 

The principle is that newcomers can benefit from the incumbent’s trailblazing to leapfrog it and claim the top slot for themselves.  

Omnity is a new search engine making use of semantic technologies that used CES 2016 to get a lot of Press coverage for its “semantic search engine” with bravado along the lines of: “We don’t view ourselves as being complementary and not competitive with Google, …. Google …is great, but no, we’re more likely to buy Google.” 

Omnity CEO, Brian Sager, talks a good game when he suggests that Omnity works in ways that  “…lets users … avoid the tyranny of taxonomy,” but he conveniently forgets that Google stopped relying on keywords in search back in 2012 and that semantic search does not rely just on clever algorithms but also machine learning breakthroughs and available data. 

Computerworld and other online publications covering it mentioned how search can deliver higher quality results when it looks at concepts and domain interrelatedness as opposed to keywords, and they’re right. The thing is that Google’s semantic search has been doing that ever since the Hummingbird recode and, in that time, it has managed to open up its lead in machine learning by at least five years from everyone else plus increase its access to data. 

An Omnity search test carried out by yours truly threw up some search results in a very pretty graphic interface showing connections but there was a distinct lack of depth and breadth and the pretty GUI is most likely going to lose its attraction after regular use when speed of access and richness of results really matter. 

Why Are Machine Learning and Data Connected?

Search is a Big Data problem and Big Data requires two things: superior, cost-effective computing power and sufficient amounts of data to train intelligent algorithms on. Machine Learning is about both of these. The new algorithms that are being created are next to useless if there is no sufficiently different computer architecture to support them and if they don’t have enough data to be trained on. 

Google has always innovated in computer architecture with the company’s engineers building their own servers from cheaply available parts since day one, pretty much. The company has also had incredible access to data through its various, popular services (Google Voice, Gmail, YouTube, Google Search, Photos and so on). If we think of the algorithms as a car engine, computer architecture is the chassis and data is the road they run on. A Formula One car with a chassis that lacks all the suitable refinements and reinforcements, running on an ordinary street road with bumps and potholes is fast reduced to the effectiveness of a horse and cart. 

That’s exactly what’s currently happening with most contenders hoping to unseat Google from search. That doesn’t mean its position is unsalable. It would only take a successful niche entry (in let’s say the entertainment and popular culture category) to create a ‘hole’ in Google’s followers by presenting superior results in that niche, to then create the precedent necessary to build a new search brand on. 

That hasn’t happened to date and since we are talking substance here, instead of style, it’s worth mentioning that Google used taxonomy (i.e. hierarchical categories strictly related by subject) to build Ontologies (lists of subjects that are related by concept) to create an index of Entities (objects and people with a distinct, independent existence characterized by a list of independently agreed upon attributes). 

Anyone who suggests that taxonomies do not matter or are restrictive is probably still thinking that Google search relies on keywords which suggests their own thinking is a little out of date, already.