Development Of An Explainable Natural Language Query Driven Data Visualization System

Abstract

Nowadays, visual interactive systems (Vis) are attracting more attention in research and industries because of their effectiveness in conveying information. Additionally, to make rational decisions based on extracted data, Vis is critical for identifying and comprehending trends, outliers, and patterns in data. Existing research has employed a broad range of methodologies to yield visualization insights into certain decision-making systems, allowing participants to perceive a specific problem from a wide range of viewpoints. However, there are still enough scopes to design a new Vis especially using visualization-oriented natural language interface (V-NLI) where state-of-the-art NLP techniques are utilized to visualize the data according to the user’s NL queries. Furthermore, in several real-life decisionmaking scenarios, this DV tools are required with proper explanations to build trust on the predictions of the model. In this regard, we propose a framework for explainable V-NLI based data visualization system. Therefore, (i) we developed a deep learning-based NLP framework to extract key information to generate proper visualization type (viz-type) on given user query. (ii) Next, we extend our prior model to an explainable visualization model that not only accurately visualizes the desired data but also explains why it appears depending on the given natural language query (NLQ).