Insights into the use of facets in search result management
The use of facets and filters to enable users to work their way through large results sets is now commonplace. As with so many aspects of search, facets and filters have been around for a long time and are a product of academic research carried out in the UK. Steve Pollitt and a team at the University of Huddersfield came up with the initial concept in 1996 and I remember being in the audience at the Online Information Conference in London when he presented the concept for the first time. There was a collective Wow! from delegates, most of whom were probably familiar with the work of S.R Ranganathan on library classification. The research baton was then taken up by Marti Hearst (at that time working at Xerox PARC) in what eventually became the Flamenco project. The history and philosophy of faceted search is recounted by Daniel Tunkelang, who has also written a good introduction to the topic.
Over the last couple of decades the focus on the effective deployment of filters and facets has been on UI development, with the Nielsen Norman Group making some very useful contributions. There has been very little research at a search interaction level, which makes a recent review paper (January 2019) on Understanding Faceted Search from Data Science and Human Factor Perspectives by Xi Niu, Xiagnyu Fan and Tao Zhang of very considerable interest. This paper appeared in ACM Transactions on Information Systems, which is available only to members of the ACM. I have not been able to find an open access version.
The authors investigated user real-time interactions with facets over the course of a search from both data science and human factor perspectives. They adopted a Random Forest (RF) model to successfully predict facet use using search dynamic variables. In addition, the RF model provided a ranking of variables by their predictive power, which suggests that the search process follows rhythmic flow of a sequence within which facet addition is mostly influenced by its immediately preceding action. In the follow-up user study, they found that participants used facets at critical points from the beginning to end of search sessions. They noted that the use of facets was different in known-object and exploratory search and was often used to exclude results that seemed to be of little relevance rather than just refining down a list of relevant results. Facets were also used to gain ideas of possible approaches to the query, and sometimes these resulted in a reformulation of the query.
As with most academic research projects the paper reports on a very small-scale experiment using undergraduates. Scaling this in a single leap to enterprise search is not sensible. However, the approach taken provides a good basis for further research. From an enterprise perspective the paper may well start a useful discussion on the extent and value of the displayed facets, which so often seem to be a function of the metadata available than the value to a user in coping with large collections of results.