Pipe(line) Dreams: Fully Automated End-to-End Analysis and Visualization

Cole Beasley, Azza Abouzied

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

We exploit large language models (LLMs) to automate the end-to-end process of descriptive analytics and visualization. A user simply declares who they are and provides their data set. Our tool LLM4Vis sets analysis goals or metrics, generates code to process and analyze the data, visualizes the results and interprets the visualization to summarize key takeaways for our user. We examine the power of LLMs in democratizing data science for the non-technical user and in handling rich, multimodal data sets. We also explore LLM4Vis's limitations, opportunities for human-in-the-loop interventions, and challenges to measuring and improving the robustness and the utility of LLM-generated end-to-end data analysis pipelines.

Original languageEnglish (US)
Title of host publicationHILDA 2024 - Workshop on Human-In-the-Loop Data Analytics Co-located with SIGMOD 2024
PublisherAssociation for Computing Machinery, Inc
ISBN (Electronic)9798400706936
DOIs
StatePublished - Jun 14 2024
Event2024 Workshop on Human-In-the-Loop Data Analytics, HILDA 2024, Co-located with SIGMOD 2024 - Santiago, Chile
Duration: Jun 14 2024 → …

Publication series

NameHILDA 2024 - Workshop on Human-In-the-Loop Data Analytics Co-located with SIGMOD 2024

Conference

Conference2024 Workshop on Human-In-the-Loop Data Analytics, HILDA 2024, Co-located with SIGMOD 2024
Country/TerritoryChile
CitySantiago
Period6/14/24 → …

ASJC Scopus subject areas

  • Computational Theory and Mathematics
  • Computer Science Applications
  • Information Systems

Fingerprint

Dive into the research topics of 'Pipe(line) Dreams: Fully Automated End-to-End Analysis and Visualization'. Together they form a unique fingerprint.

Cite this