Information Technology Reference
In-Depth Information
approach where entities are considered as documents and attribute values as
weighted terms.
At the same time, the use of faceted browsers for web search services has
become widespread. Faceted browsing allows items in a data collection to be
filtered based on the selection of values of one or more properties of these items.
For example, the Flamenco image browser [12] provides facets such as shape,
colour, location and date to search and browse image collections. Unlike a simple
hierarchical scheme, faceted browsing gives users the ability to find items based
on more than one dimension. Another example is yelp.com, a local search and
reviewing platform, where users can browse for information of interest using a
mix of keyword search and filtering. A user might initially search on keyword
and then refine the result collection based on multiple facets which are then
refined with every filter selection.
Faceted browsers come in various flavours and have been extended with var-
ious features. Facets are usually visualised as ordered lists of possible values,
where each value is followed by the number of items associated with that value.
Non-directional browsers such as Flamenco [12] and the faceted browser for
DBLP
3
offer multiple such lists, from which users can select values as part of
the filtering process. In the case of Flamenco, the selection of a value in one
facet filters the values of all other facets, but DBLP does not offer such synchro-
nisation of facets. Directional browsers, such as Apple's iTunes, have a specific
order of facets, most often represented as columns, where the browsing process
goes from left to right. The selection of a facet value in one column, triggers a
filtering action on facet values of all subsequent columns. In [11], they extend
the column representation of facets with the concept of backward highlighting,
where the selection of an item or facet value highlights all possible facet values of
precedent columns associated with the current selection and that could have led
to that selection. In [10], the information presentation is extended with so called
elastic lists
4
that visualise the weight cardinality of the facet values. Facets are
presented in the form of an ordered list where the size of a facet value indicates
the cardinality of information items associated with that value.
Both faceted browsing and tag clouds simplify search processes for users,
but have limitations in terms of how they are usually used. Often they only
support searches over one particular data collection such as products in the case
of online stores or publications in the case of DBLP. While tag clouds offer
richer visualisation in terms of being able to encode different properties of a
data collection in a single visualisation, clearly only limited information can
be visualised at one time and usually they support only very simple selection
processes. We propose an approach that combines the features of facets and
tag clouds, and extends their use to support more general search services over
multiple data collections. Our approach has been inspired by [2], where they
provide a search tool to summarise, browse and compare search results over
clinical trial data that combines faceted browsing with tag cloud visualisation.
3
http://dblp.l3s.de
4
http://well-formed-data.net/experiments/elastic_lists
