Natura: Towards Conversational Analytics for Comparing and Contrasting Time Series

Abstract

Comparing and contrasting two things is one of the most basic ideas in science. While existing tools allow such comparisons for pairs (e.g. texts, trees, graphs, histograms), there are currently no such tools for time series data. This is somewhat surprising, given the ubiquity of time series data in modern life. One could imagine a tool for comparing and contrasting two time series by reporting various summary statistics. However, scientists, engineers, and physicians typically communicate such findings in natural language. In this work we present Natura, a domain agnostic natural language framework for comparing and contrasting two time series, that aims to duplicate this human skill. Through case study, we demonstrate the effectiveness and utility of our framework.