Volume 17, Number 4

Hard-Soft Data Fusion with ChatGPT: Toward Structured Representations and Automated Reasoning

  Authors

Nicholas Gahman and Vinayak Elangovan, Penn State University Abington, USA

  Abstract

Hard and soft data fusion is a foundational concept in data science and information fusion, enabling the integration of quantitative (hard) data with qualitative (soft) information to provide richer, more actionable insights. This work investigates multiple strategies for fusing hard and soft data, analyzing their respective strengths and limitations. As a step toward systematic comparison, one such approach is implemented and evaluated. Building on the ChatIE framework by X. Wei et al., this paper introduces a ChatGPT-based extension capable of transforming unstructured natural language into structured data representations. Additionally, it presents an initial prototype of an automatic inference system designed to interpret and act upon the outputs of data fusion processes, laying the groundwork for more advanced decision-support tools.

  Keywords

Data Fusion, Graph-based Fusion, Text Transformation, Hard Data, Soft Data