FacetLens extends previous faceted systems in two ways. First, in addition to more traditional facet types such as single-value, multi-value, and hierarchical, FacetLens implements linear facets. While attribute values are intrinsically categorical, linear facets permit the visual representation of order within a facet in a way that allows data trends such as temporal relationships to be pre-served and exposed. Second, it provides users with the ability to pivot between related facets at any point during the exploration. This is important because it allows users to maintain a sense of context while they quickly and efficiently explore various areas within the dataset

Since filtering is the core and sometimes only interaction technique in traditional faceted browsers, users often find themselves unable to proceed with potentially useful exploration once they have exhausted all filters. When this occurs, they are left with a set of items, from which they must ei-ther remove filters and try new ones, or start a brand new search. This is true even though there are often complex and interesting relationships between items.

To enable users to navigate further into related items whether or not filters have been exhausted, FacetLens offers a new operation we call a pivot. Pivots are a means to reset the view to show related items, enabled by allowing items to be attributes of other items. For example, in the Papers by Author facet, author items are attributes of paper items, but each author attribute bubble also represents an author item in the dataset (Figure 1 and 2).

One set of interesting insights that can be made about a large collection of data such as a conference publication dataset involves meaningful trends. For example, which topics have come and gone in a conference? What is the publication trend of an author or an institute? What is the citation pattern of an author or a paper? Unfortunately, since the focus of most faceted systems is on presenting categorical selections, many systems also end up removing important visual ordering information. Facets are treated as lists of labels on subsets of the data, and so are attribute values within the facets.

FacetLens introduces the linear facet in which attribute values are visually presented in an explicit order, such as time. This permits exposing rich relationships between attribute values within a facet. Within the interface, linear facets are displayed horizontally across the bottom of the Facet area.

In faceted classification systems, the attributes of the items are grouped into multiple orthogonal categories called “facets.” For example, a database of books might have a Year facet to group together publications from the same year, and a Genre facet exposing attribute values like “fiction,” “his-tory,” or “education.” Providing multiple ways to reach items by presenting several facets simultaneously alleviates the drawbacks of any single categorization scheme.