Bringing Intelligence to Search
I’ve been hearing a lot lately about cognitive search, the next generation of enterprise search. It revolves around artificial intelligence (AI) applications, making search more intelligent and packing more power than a simple list of search results. Cognitive search learns from previous search queries. It knows that a search for ATM means you’re looking for a money machine, not that you’ve misspelled ATOM. But it also knows that, if you’re a physicist, you probably are looking for information on an atom, but you’re not keen on spell check and you typically forget to turn off your CAPS LOCK button. Your results are tailored appropriately.
I find it interesting that today’s highly touted AI technologies, such as natural language processing (NLP) and machine learning, aren’t all that new—computer scientists have been working with NLP since the 1950s and machine learning since the early 1960s. Yet recent advances have propelled cognitive search out of the computer scientists’ laboratories and into the business world. The explosion of data available for analysis is a key driver for the popularity and practical adoption of cognitive search.
The Case of the 11-Year-Old
If you’re looking for a simplistic prototype of how cognitive search operates, I offer this story from the 1990s. When my son was 11, he walked home from school as usual. I had a home office at the time and was on the phone with a colleague, Lynn, discussing a work project when he came in the front door. A bit later I noticed my son wandering around, clearly looking for something he’d lost. I no longer remember what it was, but it was something he’d brought home from school that he needed for homework. I said to Lynn that I was going to end the conversation and help him look for whatever it was. She said, “Tell him to look under the jacket he threw on the floor.” Now, Lynn was over a thousand miles away. She’d never met my son. She did have a son a few years older than mine. Sure enough, the object of his search was under the jacket he’d thrown on the floor. Although it didn’t dawn on me at the time, that was a metaphor for cognitive search.
In cognitive search, Lynn would be a computer, learning from many search queries about the behavior of 11-year-old boys coming home from school and dropping their jackets, backpacks, and school materials on the floor. The search engine would know where to find what the 11-year-old was searching for before he formulated the query. The computer knows that because it has knowledge gleaned from millions of similar queries from 11-year-old boys. But cognitive search goes well beyond that. It also has a knowledge base for boys of other ages and for girls. Extrapolate that one scenario into the myriad of queries from all types and ages of people and you’ve got intelligent search. It anticipates information needs based on what it knows about you and what it knows about queries from others like you. It builds on language people typically use, not just an established vocabulary. It’s smart and getting smarter.
A Case for Sherlock Holmes
Lou Jordano, chief marketing officer for Attivio, puts it well with his metaphor of cognitive search being akin to Sherlock Holmes’ “highly advanced powers of observation and reasoning.” Machine learning, he writes, enables a search system to “infer intent” from a stated query by coupling it with known personalization attributes. His point that search innovation manifests
itself in the consumer market well before it moves into the business world is very well taken. Although search has matured for individuals, within the enterprise, those innovations haven’t taken hold to the extent employees expect. Jordano pinpoints security as the culprit.
Sinequa’s CMO, Hans-Josef Jeanrond, presents a simple formula to explain cognitive search: Cognitive Search = Search + NLP + AI/ML. Analytics don’t exist in a vacuum, but ingeniously combined with existing analytical methods, they lead to better and more insightful search results. He, too, sees security as important, but it’s not a deal breaker for introducing innovative search technologies into the enterprise. Jeanrond shares two case studies—one about regulatory compliance and the other about maintaining and repairing complex systems.
Although he doesn’t stress the importance of massive amounts of data, Franz Koegl, CEO of IntraFind, does focus on the differences between structured and unstructured data. In addition to AI and machine learning, he mentions deep learning as part of cognitive search. With so many connectors existing within an enterprise, cognitive search excels at pulling together relevant information from disparate sources with a user-friendly interface. Koegl sees knowledge graphs, chatbots, automatic contract analysis, and technology scouting as exemplifying what contributes to the high-quality decision making that cognitive search encourages.
Taking a practical approach to intelligent search, Kelly Koelliker, director of solutions marketing at Verint, spells out her ten best practices for cognitive search. She’s not expecting the software to do all the work—some of the burden for making content discoverable falls on people. Artificial intelligence is great, but so is human, not so artificial, intelligence. If you take steps to make content findable, search will be much improved, as will employee efficiency.
When I think about cognitive search, it’s clear we’ve moved way beyond simply finding whatever an 11-year-old boy needs to do his homework. We’re entering a world where search is integral to knowledge management and high-performing enterprises.
Companies and Suppliers Mentioned