Machine learning is about to transform enterprise search, according to Quebec-based Coveo Solutions Inc. This week, the company launches a version of its new search technology, Coveo Reveal, specifically targeting Salesforce.
At the start of the century, semantic search was going to transform the web. Known back then as ‘web 3.0,’ it involved tagging information so that computers could understand what it meant. Fifteen years on, Coveo’s senior vice-president of products Richard Tessier believes that its time has passed.
“For semantic search to achieve good results it requires classification schemes for specific use cases — and these are time-consuming to create and maintain,” he said.
“While it is possible to have some success in niche areas, we find that it is much more efficient and effective to use machine learning based on user behavior, as it automatically takes into account the context of the individual in every scenario,” he continued.
Instead, machine learning uses pattern matching algorithms to spot trends in data. It will use the patterns that it sees to help refine its results over time, and even make predictions based on past trends.
Reveal wants to be ‘Google Now’ of enterprise search
Launched commercially last month, Reveal learns from users themselves as they search, Tessier said. The company has made a version available for Salesforce – Community Cloud Edition, designed to make searching more effective for customers and staff.
In consumer search, companies like Google have been working for a while on predicting what users need as early as possible in the search process, or even before. This has been the foundation for projects like Google Now.
This predictive search is typically based on context, including the users’ particular search history. This is also finding its way into enterprise search, Coveo said. In addition to the type of device and the location, Reveal also takes other factors into account, Tessier explained. These include the user’s job role (such as a level one, two, or three support person, or a manager, or the customer profile and purchase history if a customer is searching the system.
It will have the details of task at hand, too, pulling in data such as case information, and details of the product or project that the staff member is working on.
Coveo’s query pipelining system will understand what group a user is in, and deliver specific results for them to avoid customers seeing sensitive internal information. It will also learn based on the user’s own cultural influences, Tessier claimed.
“A successful intelligent search solution must support all of the languages and dialects your customers, partners, and employees use, and the intelligence to actually learn from all those people using search,” he said.
“It should analyze how individuals go about searching, and how successful they are in finding what they need, then automatically tune itself so the next user with the same attributes, challenges, and goals is more likely to be successful.”
Search is yet one more example of how machine learning is changing the way we do business. We have already seen it route customer calls, secure our systems, and help computers understand us more. As cloud resources make it easier to throw computing power at a problem, this resource-intensive technology is only likely to gain more traction.
