Application of full-text indexed knowledge graph in Chinese address matching for hazardous materials transportation

Address matching is one of the key technologies of origin-destination analysis of hazardous materials transportation. In order to solve the matching problem of complex non-standard Chinese addresses of hazardous material transportation, a self-expanding full-text indexed address data knowledge graph was proposed. A weighted Pinyin full-text search mechanism was incorporated to improve the matching accuracy of misspelled addresses. Combined with the online geographic searching engines, a multi-mode matching mechanism of addresses for hazardous materials was constructed, and a semi supervised matching method was performed for a small number of unresolved addresses. The simulation calculation of electronic waybill addresses for hazardous material transportation showed that the matching accuracy of extended knowledge graph reached 67.5% and the matching accuracy reached 91.57% under unsupervised multi-mode conditions. Besides, the proposed algorithm outperformed the online geographic search engines in both matching accuracy and matching precision when facing the issues of hard-to-parse addresses.