The text box. It started humbly enough, a simple UI element, a Hello World of graphical widgets that to this day still reminds us that even after the introduction of computer pointing devices, the good old keyboard (including the virtual kind) has maintained an outsized place as an input mechanism. And while natural language processing (NLP) is still considered the eventual nirvana of human to computer interaction, we by and large rely on text entry as our window into computers and the information world today.
I recall thinking about the text box when Google began its rapid ascent to the top of the information food chain. What struck me at the time was that as web sites were becoming fancier and more information-packed, Google simply placed a text box on its front page, beckoning users to merely type in something of interest. This all the while other search giants were pushing news, alerts and otherwise tipping their hand on what they felt might be of interest to the world at large. Google on the other hand simply asked the user for input, with no (apparent) presupposition.
I decided to take a look at the way back machine recently and did a quick comparison of both Yahoo! and Google as of Leap Day of 2000. Back then, both were considered search portals. However the difference in how each greeted the user was striking. The Yahoo! interface consisted of numerous links and categories of interest, including news, shopping, and various reference topics. Sure there was a small search text box as well, but there were many other things to do and the focus was unclear. Google on the other hand consisted of a single text box with two buttons.
See for yourself the images below.
Google Leap Day 2000 (with broken image link)
Fast forwarding to today, it’s much the same way in both cases, with Yahoo! having many links, news and topics of interest (albeit with updated designs) and Google keeping its spartan design. Today however, each site is likely to be classified quite differently from a positioning standpoint. Google is the undisputed king of search and Yahoo! has mostly outsourced search as a feature offering. For finding things on the Internet though, it turns out that the text box was all that was needed to get started.
I was reminded again of the power of the search text box recently while working with a customer looking to upgrade one of its customer-oriented data marts. The current system consists mostly of a multi-dimensional relational database, combined with some search capability around a smaller amount of unstructured data. They do the typical reporting and drill down that one would expect from such a system but what they also noted is the fact that one of their most used and important interface features is a search box with type-ahead capability. More of their end-users interact first and often with this simple text-box-based interface than any other. What’s even more telling is that this holds true for all of their data, both structured and unstructured. So despite having a neat relational model with SQL query capability, it’s the text box that users gravitate toward first; so much so that the application’s builders grafted it onto what might otherwise be accessed in a very different and perhaps more rigid way.
Not to be lost in all of this of course is the type-ahead or auto-suggest feature. Again, to users of Google and most modern search technology, this has become a well-known and now expected capability. If the text box is the subtle prompt from the system that says “ask me a question,” then type-ahead/auto suggest is the immediate feedback from the system that says “yes, I’m listening.”
At MarkLogic, this is not surprising. For years, these capabilities have been the core of how our customers have been getting information from data stored in the database. We often talk about search-oriented query when speaking to prospects and the ability to apply this to all data types, not just free text. This particular customer had come to the same conclusion based on their own internal customers and applied the principles, albeit with quite a bit of effort to make it work with a relational database. With MarkLogic, it’s a more natural fit for our NoSQL database with built in search and application services, providing search and type-ahead capability out of the box. So when the customer approached us, it wasn’t a surprise that they thought about how they might use our technology to not only replace what they have currently, but also to go even further and leverage our recently added Semantic capability. And though we’ve been having conversations with them about how to model things in RDF and what the SPARQL queries may look like, all of these discussions have been around the variations of how user input will be interpreted to enable richer comprehension of meaning and context.
The user interface itself however, will remain largely unchanged.
Long live the text box.