A Large Language Model-Based Database for Analyzing the Battery Critical Minerals Supply Chain South China University of Technology

Format:

Book

Conference/Event

Author/Creator:

Zhu, Juntong, author.

Luo, Wei, author.

Zhang, Xiang, author.

Yang, Zhifeng, author.

Ou, Shiqi(Shawn), author.

He, Xin, author.

Conference Name:

WCX SAE World Congress Experience (2026-04-14 : Detroit, Michigan, United States)

Language:

English

Subjects (All):

Supply chain management.

Artificial intelligence (AI).

Electric vehicles.

Logistics.

Data management.

Research and development.

Local Subjects:

Supply chain management.

Artificial intelligence (AI).

Electric vehicles.

Logistics.

Data management.

Research and development.

Physical Description:

1 online resource

Place of Publication:

Warrendale, PA SAE International 2026

Summary:

Global geopolitical volatility is recognized as a critical threat to the resilience of the electric vehicle battery supply chain. Static, manually updated databases are inadequate for capturing the sector's rapid dynamics, resulting in significant information gaps for strategic planning. To address this, an Artificial Intelligence-driven methodology is proposed for constructing a comprehensive and dynamic database. An automated pipeline was implemented. First, real-time textual data are collected from curated news and industry sources using specialized web crawlers. Then, the unstructured data obtained undergo preprocessing, including deduplication and cleansing, to ensure quality. A core innovation involves the application of Large Language Models (LLMs) for deep semantic parsing and extraction of structured information. These models are utilized to accurately identify key entitiessuch as corporations, facilities, and production capacitiesand to delineate complex multi-tier relationships spanning from raw material extraction to final distribution. The output is a structured database that provides a data-rich representation of the global supply chain. Experimental results demonstrate that the proposed semantic deduplication framework achieves a recall of 86.3% in identifying duplicate content across multilingual texts, significantly outperforming traditional methods. Through this system, over 200,000 news and industry reports have been successfully processed and structured, encompassing more than 5,000 companies worldwide. This approach highlights the transformative potential of LLMs in industrial intelligence, offering a critical tool for enhancing visibility, fostering resilience, and enabling data-driven decision-making for sustainable mobility amid global disruptions

Notes:

Vendor supplied data

Access Restriction:

Restricted for use by site license