Abstract:  The study proposes a multi- scale building fire information retrieval method for BIM data. Firstly, a comprehensive and searchable semantic graph of building fire protection is constructed by using knowledge Graph technology to improve the comprehensibility of data; then, through natural language processing technology, the query statements are segmented and semantically disambiguated, entity recognition and syntactic analysis are carried out to identify the query intention, and the logical form of query is constructed; finally, it is mapped to the query template corresponding to different scales to achieve multilevel scale information retrieval. The experimental results show that the proposed multi- scale information retrieval method can achieve 0.929 precision and 0.795 recall rate in multi- scale information retrieval, and improve the accuracy rate of 0.293 and recall rate of 0.240 in single scale information retrieval (equipment retrieval) compared with the existing keyword based methods, which promotes the optimization and upgrading of the retrieval ability of the new fire management system. 

Key words:  , fire fighting, BIM, building fire protection, information retrieval