Overview of Refinements

Refinements are a form of structured navigation that simplifies locating specific search results by dividing your search results into logical groups, known as refinement units, based on the value of one or more of the fields that those search results contain. Refinements are complementary to clustering: clustering groups related search results together into clusters based on semantic analysis of those results, and refinement enables you to organize the current set of search results based on the contents of a specified field. Clustering works with an existing set of search results, while refinement lets you further constrain the set of search results that you are viewing and that are being clustered. Clustering does not change query results, it merely groups them together. Therefore, when you restrict search results though refinement and the new, restrictive query is submitted, the clusters for the refined search results are automatically recalculated and recreated based on the set of results that the query is being limited to.

For example, suppose that you are building an application for real estate listings. Standard queries might include a specific set of criteria such as location, number of bedrooms, air-conditioning, fireplaces, and so on. If your search collection contains pricing information as the contents of a separate field, you can support searches based on the initial criteria and can then use refinement to group search results into sub-categories (refinement units) by price range. The users of your application can then explore and refine the results of their queries by exploring either graphically or in text form.

Note: The name of the underlying technology used in Watson™ Explorer Engine for structured navigation techniques such as refinements, dashboards, and so on, is known as binning. You will encounter the term binning throughout this and other tutorials but that term is not synonymous with structured navigation or refinements.

Watson Explorer Engine supports two types of refinements: tree refinements, which create a text-only, hierarchical set of refinement units that you can select, and graphical refinements, which create a histogram and associated sliders that enable you to use a mouse to select a contiguous range of refinement units within the histogram and explore their contents.

Refinements can be delivered at the collection level (as described in this tutorial) or at the project level for federated sources. Project level refinements create groups based on the returned results (not a fast-indexed content), this means that the refinements will be specific to the query and not the entire data set. However, project level refinements are useful when the data set cannot be indexed (for example, when using an external source such as a federated search). Project level and search collection level refinements are not designed to be used at the same time. Contact product support if your deployment has this requirement.

Details for configuring project level refinements can be found in Adding Refinements for Federated Sources.

Note: The outline provided in the following section is intended as a general checklist, not as a complete list of steps for this task. The complete list of steps are in the tutorial section following the overview.