建筑领域的命名实体识别研究

Research on Named Entity Recognition in Architecture-related Fields

命名实体识别(Named Entity Recognition,NER)可以从建筑领域的非结构化文本中抽取结构化的命名实体,其应用可以为建筑行业的不同决策者提供便利。近年来,NER的应用十分广泛,但在建筑领域NER的应用研究仍处于初期发展阶段。通过整理汇总近九年建筑领域NER的文献,用图表和Python生成的词云图两种形式对国内外的研究文献进行梳理。介绍NER在建筑全生命周期的应用,重点从建筑领域相关的数据获取方式、实体的标注方法进行总结。此外,还对目前建筑领域NER所使用的自主构建的数据集和研究的模型方法进行汇总。最后,对建筑领域的命名实体识别所存在的问题进行讨论,从数据集、建筑行业、建筑全生命周期三个方面提出问题并对其提出未来的展望。

Named Entity Recognition(NER)can extract structured named entities from unstructured text in the construction field,and its application can provide convenience for different decision makers in the construction industry.In recent years,NER has been widely used,but the application research of NER in the field of architecture is still in the early stage of development.By sorting and summarizing the literature of NER in the field of architecture in the past nine years,the research literature at home and abroad is sorted out in two forms:charts and word cloud graphs generated by Python.This paper introduces the application of NER in the whole life cycle of buildings,focusing on the data acquisition methods and entity labeling methods related to the construction field.In addition,a summary of self-constructed datasets and researched model methods currently used in NER in the field of architecture is presented.Finally,the problems existing in named entity recognition in the construction field are discussed,and the problems and future prospects are proposed from three aspects:data sets,construction industry,and the whole life cycle of buildings.