Increasingly, our uniform ranks of ordered blue links are being infiltrated by rich snippets and structured results that dig deeper into the data so users don't have to. Google made this mainstream by embedding maps, images, stock charts, and more into its results. Google also experiments perpetually with options like timeline view and with new projects like Google Public Data, which makes it easy to find and visually compare statistical data.
What's the optimal format for this type of result? Will a classic snippet suffice, or is there a better way? Can we summarize or surface data with an image or chart? Can we subtract clicks by adding answers? These are the questions we must ask. Of course, users' queries are important, too, since the desired output may depend upon structured input. It's a model that only works well when we can reliably infer user intent.
It's also a pattern that challenges boundaries. When is search not a search? When it's a calculator or a dictionary or an application we have yet to define. Wolfram Alpha strays well outside the box. Natural language queries and curated data are subjected to linguistic analysis and computation. Results include tables, charts, formulas, visualizations, and dynamic controls. It's not really a search engine at all; it's a computational knowledge engine. Its results are meant to be answers. But where do we draw the line?
On the search side, subtle visualizations invite attention and analysis. At Newssift, shown in Figure 4-52, emphasis is placed on identifying patterns, trends, and relationships. It's about the discovery of meaning. Structure isn't a function of individual answers; rather, it's a way to understand and manage large sets of results.
Meanwhile, the distinctions between analytics, business intelligence, decision support, text mining, and exploratory search continue to blur. Powerful queries and visualizations that were limited to structured data are being applied to multiformat collections and unstructured text. Endeca, for instance, is pioneering an approach called guided summarization (shown in Figure 4-53), which integrates faceted navigation with dynamic visualizations that engage users in a rich dialog with their data. While search may still start in a box, the potential applications of modern tools are all over the map.
In summary, structure is reshaping our results. Search applications are simply better when they swap a picture for a thousand words or take users one step deeper into the data. And this pattern plays well with others. While rich snippets may not appear in autocomplete, they're often part of best first. Features like movies, maps, and weather surely benefit from personalization, and faceted navigation offers a good model for managing text as structured data. Of course, when the response is an answer or a new frame for our question, we must wonder: is this really what we talk about when we talk about search?
