Abstract

Regulatory clauses are the knowledge basis for each stage of construction projects. How to automatically construct a formal knowledge graph with the help of computers and automatic rule checking is of great significance to the construction industry. Named entity recognition refers to extracting entities with specific meanings from a piece of text. It is an important part of extracting standard information for design compliance review and constructing knowledge graphs. This paper defines ten types of entity labels for building design codes, and proposes a method for identifying named entities based on deep learning in the regulatory clauses. Experimental results show that this method is accurate and effective, and has certain advantages compared with traditional statistical-based methods.

Download paper here

Download preprint here

Financial Sources: