Design of Pre-programming Library Matching System Based on Two-way Recall of Elemental Keywords and Semantic Similarity
Since the semantic similarity is not taken into account in the traditional text matching method based on Best Matching 25(BM25),a plan library matching system based on the two-way recall of elemental keywords and semantic similarity is proposed.Firstly,key elements,such as the application scope of each plan type,are extracted,and then semantic vectors are extracted from the key elements with the vector index library constructed offline.Secondly,key information,such as the types and the location jurisdictions of the emergencies,are automatically extracted,and the candidate set is obtained by keyword recall and semantic vector retrieval recall respectively.Finally,the results of the two recalls are aggregated.The optimal match is returned through scoring their similarity and sorting by the sorting module.The experimental results show that the two-way recall method based on keywords and semantic similarity can recall the relevant plans more effectively and comprehensively,enabling the fast and precise matching of emergency plans.
NETL
NSTL
万方数据
