How machine learning is taking over to make enterprise search smart

Claiming it will make finding business documents as easy and as accurate as using Google, a startup called Highspot launched on Tuesday. In fact, Highspot’s technology relies heavily on Google-like machine learning and graph analysis techniques to work its magic — something that is becoming the norm for companies that want to reinvent the ways in which workers find the people and things they need to know.

Highspot’s technology is like an enterprise social platform, a la Yammer or Jive, only it’s focused on helping people find the corporate content they need rather than on helping people communicate. Still, people are an integral part of how Highspot’s product does what it does, Chief Scientist Paul Viola explained. When it’s auto-completing search queries or displaying the results from them, the way that users interact with other users around the same content, as well as similarities in their behavior on the platform, help influence what Highspot shows and where.

Viola, who was previously¬†distinguished engineer and general manager of the¬†Relevance and Revenue team for Microsoft Bing,¬†described the technology as machine-learning algorithms on top of a knowledge graph, a description that invokes thoughts of web search engines. Google, for example, used to rely on its PageRank system for ranking search results based on the number and quality of links coming to a particular page. Now, PageRank is just part of a larger ranking system that invokes many other factors that influence a page’s placement among search results.

“At the very meta level, we’re trying to build a site that you can instrument in a way we see every major player on the web instrumenting [their sites],” Viola said.

An example of how Highspot predicts content based on who interacts with whom.

An example of how Highspot predicts content based on who interacts with whom.

Highspot is just one of a handful of companies now using machine learning and artificial intelligence techniques to try and improve search within companies. On our Structure Show podcast last week, we spoke with Ramona Pierson, whose startup Declara is trying to connect people with content they need to see and other people they need to meet across large national organizations. We’ve also covered a company called BrainSpace (previously PureDiscovery), which is working on something similar and also targeting developers via a BrainSpace API.

Actually, the older, larger players in the enterprise social networking space, such as Yammer and Jive, are also actively engaged in machine learning efforts in order to make their products that much smarter.

It’s all part of a push to inject intelligence into everything — from social networks to gas turbines — by collecting as much data as possible about how people and systems work. That’s one of the big themes of our Structure Data conference that takes place Wednesday and Thursday in New York, and features dozens of speakers talking about how some of the biggest and smartest companies around rely on data to deliver consistently better products and even build new ones.

Today it’s enterprise search and Facebook, but when truly smart homes and self-driving cars become commonplace, the brains that power them will be built on predictive algorithms, too.