Hierarchical Indexing for Interactive Zooming of Document Clusters
摘要
The increasing availability of large collections of textual data online gives rise to the challenge of developing methods to explore and organize large document collections in a meaningful and pragmatic way. Thematic exploration solutions allow users and organizations to effectively visualize document clusters and browse them thematically or topically. Existing solutions provide methods to topically visualize document clusters. However, they do not allow the user to dynamically explore the clusters online by zooming in and out of the cluster hierarchy. Also, most studies do not provide a solution for persisting the document clusters offline to be re-used and visualized at a later time. In this study, we design and implement a hierarchical indexing solution that allows dynamic and interactive zooming of document clusters. Our solution consists of two major components: 1) a dynamic indexing component that produces a topical and hierarchical index for document clusters, and 2) a visualization component that provides interactive and thematic exploration functionalities. The index is dynamically updated online using incremental clustering following the user’s exploration. Users can seamlessly zoom in and out of clusters to explore topically related data. Our solution serves as a baseline for an interactive cluster zooming solution and can be extended to support semantic zooming and more advanced querying functionalities at a later stage.