Semantic Word Clouds with Background Corpus Normalization and t-distributed Stochastic Neighbor Embedding

Abstract: Many word clouds provide no semantics to the word placement, but use a random layout optimized solely for aesthetic purposes. We propose a novel approach to model word significance and word affinity within a document, and in comparison to a large background corpus. We demonstrate its usefulness for generating more meaningful word clouds as a visual summary of a given document. We then select keywords based on their significance and construct the word cloud based on the derived affinity. Based on a modified t-distributed stochastic neighbor embedding (t-SNE), we generate a semantic word placement. For words that cooccur significantly, we include edges, and cluster the words according to their cooccurrence. For this we designed a scalable and memory-efficient sketch-based approach usable on commodity hardware to aggregate the required corpus statistics needed for normalization, and for identifying keywords as well as significant cooccurences. We empirically validate our approch using a large Wikipedia corpus.