An Investigation into Information Navigation via Diverse Keyword-based Facets

Abstract

In the age of information overload, it is necessary to provide effective information navigation tools that operate over unstructured textual data. Current state-of-the-art methods are limited in terms of providing effective exploration capabilities for various information seeking tasks, especially those arising in domains such as online journalism. Here we argue for improvements in faceted search systems, via new strategies for identifying keyword-based facets. Our proposed technique utilises a PageRank model operating over the graph of terms appearing in documents, while also employing novel methods for biasing significant terms and named entities. In addition, we consider the notion of diversity within extracted keywords in an effort to maximize coverage over a range of topics. We perform experimental evaluations over issues relevant to the Irish General Elections 2016, demonstrating the superior performance of our proposed technique.