Multivariate datasets with many variables are increasingly common in many application areas. Most methods approach multivariate data from a singular perspective. Subspace analysis techniques, on the other hand. provide the user a set of subspaces which can be used to view the data from multiple perspectives. However, many subspace analysis methods produce a huge amount of subspaces, a number of which are usually redundant. The enormity of the number of subspaces can be overwhelming to analysts, making it difficult for them to find informative patterns in the data. In this article, we propose a new paradigm that constructs semantically consistent subspaces. These subspaces can then be expanded into more general subspaces by ways of conventional techniques. Our framework uses the labels/meta-data of a dataset to learn the semantic meanings and associations of the attributes. We employ a neural network to learn a semantic word embedding of the attributes and then divide this attribute space into semantically consistent subspaces. The user is provided with a visual analytics interface that guides the analysis process. We show via various examples that these semantic subspaces can help organize the data and guide the user in finding interesting patterns in the dataset.

Organizing multivariate data spaces by their dimensions or attributes can be a rather difficult task. Most of the work in this area focuses on the statistical aspects such as correlation clustering, dimension reduction, and the like. These methods typically produce hierarchies in which the leaf nodes are labeled by the attribute names while the inner nodes are often represented by just a statistical measure and criterion, such as a threshold. This makes them difficult to understand for mainstream users. Taxonomies in science, biology, engineering, etc. on the other hand, are easy to comprehend since they provide meaningful labels at the inner nodes as well. Labeling inner nodes of taxonomies automatically requires the identification of hypernyms. Our proposed framework, called Taxonomizer, takes a visual analytics approach to meet this challenge. It appeals to the wisdom of humans to liaise with state of the art data analytics, neural word embeddings, and lexical databases. It consists of a set of visual tools that starts out with an automatically computed hierarchy where the leaf nodes are the original data attributes, and it then allows users to sculpt high-quality taxonomies for any multivariate dataset.