A Question of Semantics

Page 1 of 2

My family and I recently hiked to the top of the Pinnacle at Steep Rock—apt name. It is a very steep rock, but the view from the top is worth the climb: a panorama of the Litchfield hills and Lake Waramaug, Connecticut’s second-largest natural lake. After the 45-minute hike, my 4-year-old did a happy dance and wondered aloud, "Can we see the university from here?" Okay, the view is good, but the nearest university is about 40 miles away and it’s doubtful she’s aware of its existence, much less what it looks like or where to look for it. Suffice it to say, this gave us pause.

Then my husband did a remarkable thing: He asked her if she meant, "Can we see the universe?" He’d automatically shifted into "did you mean?" mode, and pulled the right word out of his lexicographic hat. I’d stalled marveling at her early awareness of higher education while he actually figured out what she was getting at. I suppose he’d unconsciously made the connection because both words are derived from the Latin universitas, meaning the whole, total, world, universe/university.

This sort of agile thinking is something we do everyday. Perhaps we miss a word or two on a bad phone connection, or we don’t express our thoughts clearly in an email, yet the person on the other end of the communication detects myriad clues from context or experience to interpret meaning, or asks follow-up questions for clarification.

With my first experience chairing the Enterprise Search Summit just (as I write) completed in May and my second coming up in September (about 6 weeks after this column is published due to print lead time), I have been focused on search. One thing I’ve found is that at such targeted events it is easy to lose sight of the reality for the average worker. Some exhibitors, for example, question the number of "beginner" sessions on the program, feeling that the market has matured to the extent that everyone is looking for an enterprise solution that is way past simple text, nay, verging on artificial intelligence.

However, the beginner sessions, such as the nuts and bolts of selecting a search engine, remain the best attended. Judging by the full rooms, decision makers still struggle with how to make search tools usable, particularly as search is the de facto gateway to information in and outside an organization.

As illustrated in my own office—when I witnessed our newest editorial assistant, Eileen Mullan, put a single term into a search box—it is unrealistic to expect the end user to be trained to "search better." The tools have to do the heavy lifting and make the machinations invisible to users. I feel certain that with a year of journalism experience behind her, Eileen will become a more complex searcher, yet the search industry shouldn’t expect her to. While finding information is part of a journalist’s training, better tools enable her to focus more on finding meaning in information rather than investing time in searching or training to search faster.

One of the information industry’s great quests is for smart search, which senses what the user is really looking for and delivers it like magic. More practical routes lead users to what they search for through navigation, clarifying terms, facets, questions, and so on. But I admit it: I like magic. Even if I know that it isn’t real, the illusion of something simply appearing is wonderful. No doubt search is hard, but epic advanced-search interfaces are certainly not the solution.

Since magic isn’t likely to become a solution anytime soon, some think semantic search may be the answer, at least for relational search. Rather than using page rank algorithms, semantic search uses—you guessed it—semantics to produce more relevant results: Does "bark" refer to doggy communication or the surface of a tree? In theory, a semantic engine would "know" that I’ve been hiking, but I wonder if a recent concern about the volume of a canine companion would confuse the engine?

I have peeked at a couple of semantic search applications, such as AskWiki, and I found that they often required more than a single-word search query to achieve the goal of providing "better" answers, which returns me to my qualms about training searchers. Some rely on the natural language approach—phrasing a query as a question, for example—which is a lot better than intricate advanced interfaces. Another tactic, used by IBM’s OmniFind email search, requires users to create "concepts" up front so that a basic keyword search will turn up what a user is really looking for. While the average worker is unlikely to invest this sort of time in setup, some would, and there are organizations that would invest IT resources when deploying certain systems in order to minimize search time for common queries.

While semantic search may not be the search industry’s magic bullet, the exploration of various routes will help users better find their way to the information, and I, for one, am enjoying the view. 

--  Michelle Manafy
    Editorial Director, ITI Enterprise Group

Page 1 of 2