Knowledge graph construction method of industrial private data based on ALBERT-XL
HU Di
WU Xingchen
LIU Xiaopeng
WU Zhitao
Abstract：This paper studies the construction method of industrial private data knowledge graph.In the indexing stage,aiming at the spatio-temporal complexity caused by small samples and fragmentation of industrial private data,this paper proposes to use the preprocessed general knowledge graph structure information to pre-train and tune the encoder attention mechanism of the ALBERT model,and integrate the Transformer-XL framework into the ALBERT Transformer structure to improve its long text processing ability.The results show that the GraphRAG ALBERT-XL model can effectively generate a knowledge graph of industrial private data with high content fidelity,which solves the problems of high spatiotemporal complexity,lack of domain knowledge,and difficulty in long-distance dependent modeling of traditional methods,and experimentally verifies its efficiency and accuracy in data processing in the smelting and processing industry.
Keywords：industrial private dataGraphRAGknowledge graphALBERT
Publication Date：2025-07-31
Pages：5( 10-14 )