Thursday, April 12, 2012

Faceted Search vs. Faceted Browse

If you have considered different kinds of taxonomies, you have undoubtedly come across the faceted type.  You can remember what a facet is by thinking of “face,” as in a multi-faceted diamond.  Other names for facet include dimension, aspect, or attribute. It could be the set of characteristics that describe a product (category, size, color, price, intended user, etc.), an image (thing, persons, location, occasion, etc.), or a document (document type, topic, author, source, etc.). In a business or enterprise taxonomy, facets for content management may include content type, product or service line, department or function, and topic. Named entities, such as person names, company names, agency names, and names of laws might also each be a facet. Facets allow users to limit, restrict, or filter results by chosen criteria, one from each facet, that are combined in any order.

Are “faceted browse” and “faceted search” the same? These designations are often used interchangeably, and until recently I had not considered a difference, preferring to use the terminology of my client. Yet “browse” and “search” are clearly not the same thing. To browse is to skim or scan a displayed list of taxonomy terms, whether arranged alphabetically, hierarchically, or a combination. To search is to enter search terms into a search box (which may then be matched against a controlled vocabulary for more accurate results). The implementations of facets in a user interface vary greatly, so perhaps the different designations of “faceted browse” and “faceted search” should reflect these different implementations.

One implementation of facets is to allow the user to dynamically restrict, filter, or limit a data set , based on selecting values from each of multiple facets that are displayed, typically in the left-hand margin, while references to the data or content is displayed in the main screen area. Under each named facet are displayed the names of values (taxonomy terms) within the facet.  Facets may need to be expanded to display all values under each, or there may be scroll bars of terms. This implementation of facets can be considered “browse” because the user browses the displayed facets and the displayed terms within each facet.

The data set that is filtered by the facets could be the entire set of content, but more likely it is a subset, based on a prior execution of either a category selection or a search.  If the user’s first step was to initiate a search to obtain search results, and then uses facets to limit the search results, this might be called “faceted search.” Even though the user browses the facets, because the facets are introduced as a second step following search, this step might be called “faceted search.” If, however, the user’s first step was to browse subject categories and select a category to obtain the initial data set, then the use of facets in the second step would more likely be called “faceted browse.” I would consider it better practice to call the process “faceted browse” in either case, regardless of how the initial data set was obtained. However, if it’s less confusing to the users, I will defer to those who prefer to call this process “faceted search.”

Another implementation of facets is to allow the user to select among limiting criteria from the beginning, without first selecting a subject by browse or search. In order to achieve usable results (result sets that are not too large), the facets need to contain relatively large taxonomies: a large number and deep set of terms. While it is certainly possible to display a large taxonomy for browsing, it may be difficult to display multiple large, browsable taxonomies, one for each facet. Therefore, if facets are made available to the user from the start (without first requiring the user to select a limited data set based on a search or browse selection), it is more likely that that not all the facets will display the terms to the user. The user must then execute a search within a facet. This would correctly be called “faceted search.”  It is also known as “fielded search” or “advanced search,” as a search field/box is made available for each facet “field.”

The distinction between faceted browse and faceted search is lost, however, where the distinction between browsing and searching is becoming blurred. Newer user interface implementations of taxonomies are combining search and browse, so that the difference is no longer as obvious. For example, I have seen cases where there is a search box, and as the user types in something, a type-ahead feature matches the search string against controlled vocabulary terms, which are displayed in a short list under the box, and the user can browse the list to select a term.  I have also seen a case where a user may be presented with a search box to enter search terms, and there is a button next to the search box, which the user may optionally click, and then the search box becomes a scroll box to view and browse the entire controlled vocabulary for that field. When these kinds of advanced taxonomy-enhanced search boxes correspond to facets, the distinction between “faceted search” and “faceted browse” truly no longer exists.