Fully Dynamic Semantic Word Clouds

Word clouds are popular graphical representations used to effectively summarize the core content of a document, or of a set of documents. In particular, in a semantic word cloud the positions of the words also reflect their semantic correlation. Several algorithms have been proposed in the literature to compute semantic word clouds, both in a static and in a dynamic setting. However, so far, solutions proposed to generate dynamic word clouds assume that all text changes are known in advance. In this paper we propose a new dynamic scenario. We assume to receive a text stream in a given time period, and we compute, in a sort of real-time fashion, a dynamic word cloud that shows the evolution of the stream. The visualization should preserve both the semantic relationships between words and the user's mental map. We present an algorithmic framework and the results of an experimental study that compares different solutions for the steps of its algorithmic pipeline.