Using Document Indexers for Faceted Search in Dataspaces

Abstract

INTRODUCTION Two main ways of information retrieval are browsing and searching. Browsing demands information seekers' attention and time, and when the tasks are repetitive, fatigue sets in quickly (Yeh & Liu, 2011). Browsing, especially through page after page of alphabetical listings with a browser, could be frustrating to users. Research has been done to study the approaches to make the browsing experience more efficient and effective. Grouping information as categories during browsing, also called faceted browsing or faceted navigation, is one of the approaches proposed for enhancing the users' browsing experience. Faceted browsing offers the users an overview of results, and narrows their list (Yeh & Liu, 2011). Since browsing naturally involves gathering data with conceptual grouping, faceted browsing has been demonstrated in facilitating more effective browsing and information retrieval, and in speeding up the users' decision making process (Kim et al., 2014). Faceted search is a natural extension to faceted browsing. Searching is only effective if the query is formulated a priori (Kim et al., 2014). When the queries are difficult to define a priori, the search results may be irrelevant and the users may experience undesirable outcomes from their direct search alone. To make direct search more efficient, faceted search has been proposed. The faceted search approach integrates faceted browsing with direct search methods (Yeh & Liu, 2011), thus assisting users determine which portions of the facet contain more relevant and desired information (Perugini & Ramakrishnan, 2003). FLexible information Access using MEtadata in Novel COmbinations (FLAMENCO) was developed as an open source search interface framework (Yee et al., 2003). It utilizes faceted navigation for both browsing and searching. Since then, faceted searching has been used in various fields and has been proven to generate better search results than direct search (Hearst, 2006). Faceted search solicits and captures search keywords, then prunes out branches of the hierarchy irrelevant to the user's information need (Perugini, 2010). A major limitation of using faceted search is that categories are predefined and fixed regardless of the difference in search results. Recently, Kim et al. (2014) used semantic web to design a dynamic faceted search system. Their results showed that using it can save considerable time for users and improve the efficiency of search. Another key problem of a faceted search system is its limitation in searching within a very large and complex document domain that typically exists in big organizations and government agencies. These organizations usually create dataspaces to integrate and manage data and documents from many loosely connected heterogeneous data repositories. Some faceted-search systems show users all available facets and facet-values. This approach can quickly overwhelm the users and lead to diminished user experience and search performance (Sinha & Karger, 2005; Koren et al., 2008). Based on the extant designs on faceted browsing, faceted search, and dynamic faceted search, our research considers the use of a complex ontology defined to support the dynamic category selection. In particular, the categories are generated for the navigation of a document with relations to subdocuments. The sample ontology (Figure 1) shows that a conference proceeding has relations to two groups of subdocuments: editors and articles. In addition, an article has relations to a group of authors. Eventually, all categories in faceted navigation may include all the properties of a proceeding and its subdocuments (such as title, year, publisher, article title, and article author's last name). A user enters keywords to form a query based on the ontology and receives a result including proceedings categorized and arranged by a faceted search system (Figure 2). The result may still include a large collection of proceedings. …