Study On The Key Technologies Of Domain Database-Oriented Question And Answering System

The QA System to database based on natural language understanding provides users with a way to access information stored in database directly in natural language.It involves many kinds of subjects,such as NLP,DB,AI and HCI,etc.Over the past forty years, although there have been signification advance in this area, the QA System to database based on natural language understanding did not gain rapid and wide commercial acceptance for the problems of portability and usability.This thesis designs a common natural language interface model to database and studies its principle, technology and solution thoroughly which is based on the establishment of a general model. Hereof this paper establishes a concrete implementation system at the same time, and hope that the system can be easily ported to other application domain.What this paper studies is to design and implement the whole theories and policies of a practical natural language interface to database.The main research results are as follows:1. At the aspect of "Knowledge Expression and Extraction", this thesis separates the establishment of knowledge_base from system programming and designs common knowledge_base and domain knowledge_base respectively which are the two major component of knowledge_base by using the ideology of "divide and rule".This system has a good performance of transplanting and flexibility.2. At the aspect of "Chinese Natrual Language Question Understanding",this thesis provides a improved MM algorithm based on partitioning search which changes the existing order matching way and heighten the efficiency of word participle and lable; The system adopts the lable method which is based on semantic database, and it changes the traditional and cumbersome procedures of lexical analysis, sentence analysis and semantic analysis. As the result of full using of the database schema information, it lightens the workload for the remaind language disposal; This thesis adopts the multi-level list as the middle language.This method is vivid and can translate between natural language and SQL; This paper puts forward the classification and identification method for every kinds of query aims and query condition, and also introduces complex concept and reason concept. This method enhances the system's ability of identify the implied knowledge in the database, and the system has better capability of vigor and intelligence.Because of what this article has adopted is a kind of binding chinese natural query language, which limit the use of structure form in case of that it can quarantee the form of chinese natural language description. Thereby, the process of natural language query interface to database becomes effective and feasible. Otherwise, if this system combines with intelligent system such as voice recognition and etc, there will be a good prospect.