Supporting Human-Centric Data Exploration Through Semantics and Natural Language Interaction

SIGMOD/PODS '25: Companion of the 2025 International Conference on Management of Data

Data science plays an increasingly central role in decision-making across domains, yet the effectiveness of these decisions hinges not only on sophisticated algorithms but also on how well systems support human interpretation, exploration, and communication of data. This tutorial explores the intersection of semantics, natural language processing (NLP), and human-computer interaction in creating human-centric data exploration tools that promote accessibility, trust, and transparency. This includes techniques for generating perceptually meaningful visual encodings, using NLP for query interpretation and ambiguity resolution, and designing conversational interfaces that align with users' intent. The tutorial will highlight research from a broad set of contributors in the SIGMOD/VLDB, HCI, and visualization communities, spanning natural language interfaces for databases, multimodal interaction systems, semantic search for data repositories, and the use of AI and large language models to augment visual and textual analysis.

As part of this 1.5-hour session, we will present case studies and systems that demonstrate how human-centric design can be integrated into the data analysis pipeline, including tools that support mixed-initiative interaction, adaptive defaults, and subjective query interpretation. We will also discuss open challenges and research opportunities, such as semantic inferencing for unstructured data, retrieval-augmented generation (RAG), and ethical considerations around fairness, explainability, and user agency. Drawing from principles in perception, linguistics, AI, and HCI, this tutorial aims to equip attendees with both conceptual frameworks and practical techniques to build more inclusive, interpretable, and intelligent data systems. It is intended for a broad audience in the SIGMOD/VLDB community interested in designing the next generation of data exploration tools that are aligned with human needs.

Tableau 작성자

Vidya Setlur

작성자

Vidya Setlur