[0006] First, there are still many problems in the application of the existing printing ERP system in actual production. For example, there are many material codes in the material management, which lead to memory difficulties, complex interfaces, slow retrieval speed, and single retrieval methods. The management system is far from being popularized in the printing industry. Regarding the problems existing in the printing industry management system and the changes that are taking place, the existing printing ERP is based on the graphical interactive interface, which is already difficult to deal with. The graphical interactive interface needs to integrate the background The storage and logical structure of data are completely presented to users, and the increase in functional coverage will increase the background data and make the database structure more complex, increase the complexity of the user interface, bring about a decline in user retrieval efficiency, and affect user experience. Enterprise management system Clients tend to be more and more convenient and miniaturized, which also makes it difficult to adapt to complex and numerous graphical interfaces. Facing the increase in intelligent interaction brought about by the transformation from enterprise management systems to business intelligence systems, the existing graphic The search interface can only handle simple search features in a single data table, which can no longer meet the needs;
[0007] Second, affected by the development of information technology and the digitalization of the printing industry, the printing industry management system has undergone tremendous changes in function and structure. The graphical human-computer interaction mode of the existing technology is becoming more and more complicated, and the interactive friendliness is declining rapidly. , NLQI can better deal with these problems, but the development of natural language processing technology in the prior art is still not mature enough. There are many difficulties in converting the natural language in which structured phrases and unstructured phrases coexist into a completely structured form of retrieval language. Natural language retrieval sentences often have complex sentence structures such as nested retrieval and combined retrieval, as well as redundant and ambiguous problems, and the database of the printing industry management system is complex, and cross-table retrieval is very common, which increases the difficulty of understanding natural language;
[0008] Third, in order to express the retrieval sentence more clearly, semantic grammar introduces semantic information into the structure of the syntax tree. Although semantic grammar has many advantages compared with the formal grammar of the prior art, the addition of semantic information also brings some questions
First of all, semantic grammar uses words with semantic information as non-completion points, which increases the degree of differentiation between non-completion points, makes the syntax tree structure more complex, and results in a substantial increase in the number of semantic grammar analysis rules, thus increasing the workload of writing analysis rules ;Secondly, the semantic information has a strong domain correlation, and the versatility of the semantic grammar will be greatly limited; in addition, there are many types of nodes in the semantic grammar tree, which makes the structure of the syntax tree extremely complex, adding subsequent semantic analysis and SQL statements The difficulty of generation even affects the accuracy of conversion of natural language retrieval sentences;
[0009] Fourth, the grammatical analysis tree essentially analyzes the sentence structure at the grammatical level, and the accuracy of the analysis is difficult to guarantee, especially because the grammatical analysis does not carry out a detailed analysis of the retrieval condition phrases and retrieval target phrases in the grammatical tree structure. Analysis, cannot guarantee the accuracy of the entity modification relationship, and has not yet been completely transformed into structured data, and SQL statement conversion cannot be performed, so semantic analysis of the retrieval conditions and retrieval targets is also required
However, as a subset of natural language, natural language retrieval sentences are highly colloquial, and unstructured phrases often exist. There is no suitable processing method in the existing technology. The construction of the parsing tree is an analysis process at the grammatical level, and there are certain limitations. one-sided problem
In addition, the natural language ambiguity phenomenon of redundant reference exists in the structured single retrieval condition, the form of the single retrieval condition is not uniform, the number of templates in the template matching process is large, the structure of the template library is complex, and the efficiency of template matching is low. Flexible handling of the ambiguity of redundant reference