Relational keyword search (R-KWS) systems provide users with a convenient means in relational database queries. There exist two main types of R-KWS: those based on Data Graphs and those based on Schema Graphs. In this paper, we focus on the latter, R-KWS based on Schema Graphs. Most existing methods are typically inefficient due to the large number of repetitive operation caused by overlapping candidate networks and the execution of lots of complex queries. We present the R-KWS approach based on combined candidate network evaluation to improve the query efficiency. The proposed Combined Candidate Network (CCN) can efficiently share the overlapping part between candidate networks, and then avoid the repetitive operation during the evaluation of candidate networks. Meanwhile, CCN possesses another important characteristic that candidate networks within a CCN are still identifiable after candidate networks being compressed into a CCN. We design an algorithm based on this characteristic to evaluate CCN for the generation of final query results. This algorithm is able to eliminate the execution of a large number of complex queries required by most existing approaches, and thus significantly improve the efficiency of keyword search. Experiments on real datasets show that our approach can improve query efficiency without any loss of the quality of query results with respect to existing approaches. INDEX TERMS Relational database, keyword query, schema graphs, combined candidate network. I. INTRODUCTION Structured Query Language (SQL) is the main means to access data from a relational database, which requires users to understand the complex SQL syntax and schema information of database [4], [27], [29], [31]. Compared with SQL, relational database keyword search (R-KWS) is simpler, which enables users to query information from a relational database by the way of search engines. Recently, lots of works about R-KWS are proposed, which can be classified into two main types: R-KWS based on Data Graphs [4]-[7], [28], [30] and R-KWS based on Schema Graphs [8]-[24]. R-KWS based on Data Graphs typically needs to preload the Data Graphs into memory, which cannot sometimes be completed at a time because of the consumption of a large amount of memory [7]. R-KWS based The associate editor coordinating the review of this manuscript and approving it for publication was Yongqiang Zhao .

At present, the uneven distribution of entities and the low frequency of some entities in medical text data leads to the low accuracy of medical named entity recognition. To solve the above problems, a neural network model based on dictionary and mutual attention (DB-MA-BiLSTM-CRF) is proposed. The model includes Bert embedding layer, BiLSTM-CNN layer, mutual attention layer and CRF layer. The medical dictionary is fused in the Bert embedding layer to inject medical vocabulary information; Then it is input into the BiLSTM-CNN network layer to extract the global and local features of the text respectively; The features extracted from BiLSTM-CNN network layer are spliced into the mutual attention layer for weighted extraction of important features. Compared with the two benchmark models, the experimental results show that the model proposed in this paper has strong advantages.

No abstract