基于深度学习和模型训练的建筑物资库数据处理方法

Data Processing Method for Building Material Repository Based on Deep Learning and Model Training

在本论文中,我们探讨了一种基于深度学习和生成式预训练模型(GPT)的建筑物资库及非结构化数据的数据智能处理方法。研究目的是提高建筑物资数据管理的效率和准确性,通过自动化处理,降低人工操作的复杂度,并为企业节省时间和成本。我们采用了多种技术手段,包括词嵌入、TF-IDF、余弦相似度、命名实体识别(NER)以及基于GPT的模糊匹配。实验结果表明,该方法在物资数据清洗、别名匹配、信息提取和表单生成等方面表现出色,具有较高的准确性和效率。此研究为建筑行业提供了一个创新的数据处理解决方案,有望在工程项目管理中发挥重要作用。

In this paper,we explore an intelligent data processing method for construction material databases and unstructured data based on deep learning and Generative Pre-trained Transformers(GPT).The research aims to enhance the efficiency and accuracy of construction material data management,reduce the complexity of manual operations through automated processing,and save time and costs for enterprises.Various techniques were employed,including word embeddings,TF-IDF,cosine similarity,Named Entity Recognition(NER),and GPT-based fuzzy matching.Our experimental results demonstrate the outstanding performance of this method in material data cleansing,alias matching,information extraction,and form generation,exhibiting high accuracy and efficiency.This study presents an innovative data processing solution for the construction industry,with the potential to play a significant role in engineering project management.