Research On The Implicit Knowledge Query Of Database Based On Plan Recognition

Natural language interfaces to databases (NLIDBs) is a method to allow users accessto database by using natural language. The appearance of NLIDBs makes the process ofhuman computer interaction much easier. NLIDBs is a product of interdisciplinary. Theresearches of natural language Understanding, database technology, Artificial Intelligenceand Human Computer Interaction are involved. With the advent of Things and thedevelopment of voice recognition technology, the study of NLIDBs is with great theoreticaland practical value.In the research of NLIDBs user interface, the key point is the understanding of thenatural language query which inputted by user. The recognitions of query condition, querytarget and query entity are known as a top priority. In this document, plan recognition andhierarchical planning ideas are used in understanding of natural language query. Use themature theoretical foundation and techniques of plan recognition in natural languageunderstanding field, to understand natural language query. It improves the intelligence andreasoning ability of database query interface.Research topics in this document:(1) Discuss about the problems in Chinese query database systems. Based on thedifferences of data representation and natural language query semantic models, databasequery language query with implicit knowledge is proposed. At the same time, relatedknowledge is provided. The ability of process query with implicit knowledge in database isimproved as focus in this document.(2) The natural language database query interface based on plan recognitions is built.Plan recognizer consists of two core modules: intelligent word segmentation and inferenceengine. By the reason of difference between Chinese and English, word segmentation isrequired for understanding of Chinese query. A new word segment algorithm is used in thisdocument: iterative maximum matching algorithm. This algorithm overcomes thedisadvantage of mechanical word segmentation to ensure that the word segmentationprocess to maximize the retention of complete words and semantic relations between wordsand implement intelligent word segmentation.(3) Combine the thinking of plan recognition and hierarchical planning and theproperty of SQL query structure, a new structure is defined in this document: planningknowledge tree. The sub-trees represent for query action, query target, query entity and query condition. The structure of planning knowledge tree not only keeps the thinking ofplan recognition and hierarchical planning, but also has the represent method of databasewith implicit knowledge. It reached the inner unity. The recognition of query action, querytarget, query entity and query condition transfers into the corresponding sub-treeconstruction. The essence of understanding process is the process of expanding and pruningof the planning knowledge tree.