According to the construction specification text
knowledge graph construction method based on the word multi-
granularity feature enhancement
algorithm, in a
named entity recognition task, a BERT pre-training
language model is introduced to serve as an embedded layer, word
granularity and word
granularity vector representation is obtained, the importance of professional words in sentences is improved while the problem that one word is
polysemy is solved, and the construction specification text
knowledge graph construction efficiency is improved. The
engineering specification entity recognition model is fused with a BiLSTM-Attention-CRF model, and an
engineering specification entity recognition model is constructed; and in the relation extraction task, constructing an
engineering specification relation extraction model by adopting BiLSTM-Attention. In combination with the engineering specification
knowledge graph, through
global information display and specific
information retrieval based on the knowledge graph, the utilization efficiency of engineering specification knowledge is improved, and engineering site construction is assisted. According to the method, the characteristics of engineering construction specification management, a
text processing technology and a
deep learning method are combined, the deep analysis of the construction specification text is realized, the mining efficiency of safety management information is improved, the engineering specification knowledge graph is constructed, and important support is provided for
intelligent management and control in the engineering field.