AstroDB Data Extraction: A Comparative Analysis of Embedding, Query Expansion, and Filtering Techniques
Year: 2025
Type: dissertation
Abstract: Due to the rapid growth of astrophysical scientific papers, it is difficult to extract important data manually from the scientific papers. While traditional search engines can help to find relevant publications, they may not be able to capture the numerical and contextual details of significance. This thesis explores the multi-modal data extraction in the AstroDB. The accuracy, LLM token usage, and run time of various combination extraction methodologies are compared in this study to reveal key precision versus computational efficiency trade-offs. These insights provide a way to improve the data extraction pipeline and accelerate scientific discovery in astrophysics.
Authors: Feng, Baoying, Feng, Baoying
Language: English
Topic: Topic Modeling
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W7064078568