基于人工智能的数据表间关联关系自动识别技术研究

打开文本图片集

中图分类号：TP391 文献标识码：A文章编号：1001-5922（2025）06-0151-04

Abstract：Aiming at the problems of difficult to identify the relationship between tables and low recognition accura⁃ cy. An automatic identification scheme of correlation between data tables based on artificial intelligence is de⁃ signed. Firstly，the data table is preprocessed by Fourier transform technology，and the structure of the data table is identified by HR-net neural network. Then，the BERT model is used to extract the text and semantics of the table text，and the LSTM model with the multi-head attention mechanism is used to summarize the semantics of each ta⁃ ble. Finally，the dynamic association network is used to realize the general semantic comparison between different data tables，and the automatic identification of the association relationship between tables is completed. Experi⁃ ments show that the scheme has strong performance in text recognition，semantic prediction and automatic recogni⁃ tion of inter-table relations.

Key words：BERT model；multiple attention；fourier spectrum transformation；relationships between data tables； automatic recognition

当今数字时代，面对海量繁复、看似毫无逻辑相关的表格，将非连续、繁琐的表格碎片信息快速有效地结合起来，是一项人力难以企及的工作[1]。（剩余4829字）