We found a match
Your institution may have access to this item. Find your institution then sign in to continue.
- Title
建筑安全事故通告关键信息自动提取方法.
- Authors
董国鹏; 徐旭升
- Abstract
Efficient safety management is an important issue in construction projects. Usually, information about safety-related accidents or incidents is recorded in construction reports, which are obscure to understand and analyze for engineering managers. With the rapid development of machine learning, researchers have made a progress on safety report analysis based on the natural language process (NLP) technique. NLP utilizes computers to extract the key words of safety reports, and thus is fairly automatic and efficient. However, the effectiveness of NLP relies heavily on the sufficiency of training data, which is usually quite limited in a construct project. In this regard, a small-dataset-based NLP approach for safety report analysis was developed. First, a text data augmentation algorithm based on cross combination was used to enlarge the dataset. Then, the “character” was used as the basic detection unit for Chinese character semantic encoding, and the bidirectional, long short-term memory, conditional random field (BILSTM-CRF) model was used as the detection kernel to process the text. The proposed approach was verified and validated in a real construction project, and is anticipated to be a promising tool for efficient safety management.
- Publication
Science Technology & Engineering, 2022, Vol 22, Issue 10, p4026
- ISSN
1671-1815
- Publication type
Article