SCIENCE@home

Contrasting Traditional and LLMs-Based Approaches For Holistic Analysis of Disparate Data Types

Abhishek Santra,Sharma Chakravarthy

Published 2025 in Proceedings of the 2025 International Conference on Artificial Intelligence, Big Data, Computing and Data Communication Systems

ABSTRACT

Research has addressed the modeling and analysis of individual data types fairly successfully whether it is mining, computing objectives, or knowledge discovery. However, handling disparate data types holistically is challenging, and can benefit from new models and/or approaches. It is also important to note that holistic analysis is significantly different from traditional data mining/analysis, which deals with a specific type of knowledge discovery using specific algorithms (e.g., clustering, association rule mining, etc.) on specific data types and formats. In contrast, holistic analysis is a broader umbrella concept and is likely to need an ensemble of modeling and analysis alternatives. Hence, for the modeling and analysis of multiple data types with different characteristics, a suite of existing approaches and their extensions, as well as new ones, is required. When dealing with diverse complex data, all aspects – data preparation, modeling, analysis, drill-down, and visualization – become important for understanding the data as well as the knowledge discovered/objectives computed. In this position paper, we present our vision for the holistic modeling of data of different types from different sources to perform analysis and knowledge discovery. This can be seen as the first step towards big data analysis. The goal is to accommodate multiple data types (structured, unstructured, image/video, and text from natural language) using models that are amenable to expressive analysis and/or infer knowledge using scalable approaches such as LLMs (Large Language Models). In this paper, we contrast two significantly different approaches: i) the use of graph and multilayer network (MLN) data models (termed the traditional approach) and the use of LLMs (termed the AI-induced approach). After motivation and the current state-of-the-art, we elaborate on the details of the traditional approach, which we understand better, and contrast it with the AI-induced LLM-based approach. We highlight the challenges and, through two different representative applications, show how these two proposed approaches can provide two different (and even orthogonal) paths towards information integration/fusion and concomitant holistic analysis.

PUBLICATION RECORD

  • Publication year

    2025

  • Venue

    Proceedings of the 2025 International Conference on Artificial Intelligence, Big Data, Computing and Data Communication Systems

  • Publication date

    2025-11-25

  • Fields of study

    Not labeled

  • Identifiers
    DOI 10.1145/3759023.3759115
  • External record

    Open on Semantic Scholar

  • Source metadata

    Semantic Scholar

CITATION MAP

EXTRACTION MAP

CLAIMS

  • No claims are published for this paper.

CONCEPTS

  • No concepts are published for this paper.

REFERENCES

Showing 1-39 of 39 references · Page 1 of 1

CITED BY

  • No citing papers are available for this paper.

Showing 0-0 of 0 citing papers · Page 1 of 1