Key points are not available for this paper at this time.
When bidding on construction projects, contractors need to understand the specifications properly to manage project risks. However, specifications are mainly analyzed based on human cognitive abilities, which can take considerable time and can lead to errors due to misunderstanding. While efforts have been made to automate this process, that the existing academic efforts to automate the process have limitations. To develop an automated specification reviewing model applicable to various kinds of specifications, the authors propose information extraction frameworks consisting of five categories. In addition, a named entity recognition (NER) model was developed based on bidirectional long short-term memory architecture to extract information from text data automatically. The data set for model development comprised 56 construction specifications, which included a total of 4,659 sentences labeled according to five categories of information. Word2Vec was utilized to aconvert labeled text data to the form of numeric vectors to be input into the NER model. The NER model successfully assigned every word in the testing data to an appropriate category with a satisfactory performance of 0.919 precision and 0.914 recall. These results contribute to the automation of the construction specification review process. Although this research focused on road construction projects, the proposed information extraction framework can be applied to other types of construction projects.
Moon et al. (Sat,) studied this question.