Bridgeline: Faceted search "is arguably the most significant search innovation of the past decade." You write: "In keeping with the principles of progressive disclosure and incremental construction, users can formulate the equivalent of a sophisticated Boolean query by taking a series of small, simple steps. Faceted navigation addresses the universal need to narrow.... Its discriminatory power to clarify intent and refine results may offset the need for personalization and advanced search." Why are facets so effective when compared to parametric search?
Peter Morville: The terms "faceted navigation" and "parametric search" are often used interchangeably, but for the sake of comparison I find it valuable to define interfaces that require the simultaneous, up-front specification of all search parameters as exemplars of parametric search. Like the Boolean queries of yore, this forces users to formulate and execute a search strategy without guidance or feedback. Sliders and pull-downs are easier than ANDs and ORs, but syntax is only part of the problem.
In contrast, faceted navigation lets users begin naturally with a keyword or two. They're rewarded with traditional results plus a list of facets (or fields) and values, usually on the left. This SERP (search engine results page) serves as a custom map that offers insights into the content and its organization. And, this is a map that's also the territory. Users can take a simple next step to clarify or refine their query by clicking on a facet value. And, by taking several of these simple next steps, users can construct a sophisticated, powerful query. So, not only do users find what they need, but they also learn along the way.
Bridgeline: As you say, "semantic search hasn't progressed much beyond parlor tricks." Why is it so hard to answer the query "Where is Java?" How can businesses meaningfully fill the gap while semantic search improves?
Morville: Natural language is sufficiently ambiguous that it's difficult if not impossible to derive meaning without context. If you walk up to a stranger and ask "Where is Java?" there's a good chance they'll know you're asking about the island rather than coffee or programming; unless you're in Richmond, in which case they may direct you to Java, Virginia. But they won't know why you're asking or how much you already know, so they're quite likely to answer with a question.
One of the best things we learned in library school was how to conduct structured conversations with patrons (or "users") to better understand what they want and need. These "reference interviews" feature open-ended questions. They are iterative and interactive. Both participants learn. Of course, it's hard to replicate this model online, because computers can't hold up their end of the conversation.
That's why I'm a skeptic when it comes to grand visions of The Semantic Web. In narrow domains such as medicine, we can develop thesauri (or "ontologies") that define terms precisely and map hierarchical, equivalent, and associative relationships. But these approaches simply don't scale, and they can't keep up with the rapid evolution of language and knowledge.
In my opinion, there's simply no substitute for taking a holistic approach that recognizes search as a complex adaptive system with multiple interrelated elements.
Can ontologies help? Sure. But we also need social data to drive result ranking algorithms like Google's PageRank and Flickr's Interestingness. We need conversational interfaces, better content, and more motivated creators. And, all of this work should be guided by a deeper understanding of user psychology and behavior, not to mention business goals. In short, there's no silver bullet. Search is a wicked problem that will never be solved.
Bridgeline: You point out that by comparing search terms that deliver people to your site, and the queries they perform when they get there, "we often expose gaps in marketing and design that we must address by improving both search engine optimization and the site's information architecture." Clients have come to understand that Search Engine Optimization (SEO) is an iterative, ongoing process, but the perception still exists that once an IA is in place it's locked down. How do we demonstrate that a site's ability to grow with its users search needs is just as important as an ongoing SEO campaign?
Morville: Stewart Brand has advanced "pace layering" or the idea that in any complex system, it's important that each layer can adapt at its own pace. In society, fashion and commerce change fast while government and culture move slow. In buildings, furniture and walls are temporary, whereas structure and foundation endure. An information architecture can be teased apart into many layers. The core structure might last for decades, while links and labels demand regular review.
For instance, search analytics is an activity that merits ongoing attention. What are the most popular queries on your site this week? What's changed since last week? How can you adapt? Clearly, you can update Best Bets. But perhaps the navigation system and site map and quick links also require review. Also, for sites like Etsy.com that combine taxonomy with folksonomy, there's a wonderful opportunity to create a feedback loop. Popular tags should be considered for integration into the taxonomy, to help users navigating within the site and to improve SEO.
Bridgeline: When we last talked eight years ago, you cited Amazon.com as one of the best "wow" experiences on the web: a no nonsense search interface that delivers pretty good results; an excellent subject classification scheme for topical browsing; and a blossoming participation economy. What has changed since then? What wows you today?
Morville: Eight years ago? Really? Wow! Well, I still use Amazon as an exemplar of integrated search and browse, but today I'm most interested in the strange stuff happening at the crossroads of ubiquitous computing and the Internet. Information is blurring the lines between product and service to create multi-channel, cross-platform, trans-media, physico-digital experiences. Netflix, Nike+, iTunes, and Zipcar are just a few examples. Or consider Redbox. I use my iPhone to search for a movie and then to search for a nearby kiosk that has that movie. Who could have predicted this weirdness? And what's great is that search is *the* bridge application that connects experiences across devices and channels.
I've said it before, and I'll say it again: Search is among the most disruptive innovations of our time. It influences what we buy and where we go. It shapes how we learn, what we believe, and who we know. It's a wicked problem and a radically multidisciplinary challenge. It's also a lot of fun. So let's not wait. We have both a tremendous opportunity and a responsibility to make search better. Let's get cracking!