| Natural Language Understanding has become one of the most active fields in Artificial Intelligence study. Natural language interface is one of the most hopeful applied fields in the research on Natural Language Processing. The accuracy of computer understanding query of natural language is key to the quality of the natural language interface. Basing on the concept of constructing a universal model, this paper introduces the design of a universal system model about database natural language interface and the principle of realizing it. A specific system has been designed and is expected to be easy to transplant to the other applied fields. The research in this paper is about the theory and strategy of the whole process of designing and realizing a practical database natural language interface system and results the following mainly:1. A universal system model about database natural language interface, which separates the programming from Knowledge-Base is constructed. In this model the computer understanding of Chinese natural query language and knowledge database are designed respectively, that is to make them independent of each other. There are modules of lexical analysis, syntactic analysis, semantic analysis, natural language translating middle language, middle language translating computer language and making response in Understanding and these modules are independent relatively. The design makes the system separate from specific applying fields, and so become independent, which is convenient for transplanting and extending.2. It shows that the Knowledge-Base is divided into three layers: the natural language understanding Knowledge-Base, the Chinese query sentence understanding Knowledge-Base and the application Knowledge-Base, and each layer should include its Knowledge-Base and the structure of the Knowledge-Base.3. A Chinese-natural-language-words-cut and word-tagging system software is designed and programmed, and this subsystem uses statistical information andgrammatical rules. Much attention is paid to syncopate ambiguity finding, disambiguating, name recognition. Firstly, it provides the max-length words and second-length words match basing on lexicon with maximum matching method. Then, it checkes whether there are ambiguities or not . Then, it judges the ambiguous type. Finally, it raises a set of methods to disambiguate according to the different ambiguous type. The system can load the user's lexicon and it can excel the normal lexicon, and can be used for all kinds of subjects' sentences and words cut without any modifying.4. A medium language, which is not only fit to query language semantic representation but also better to be changed into computer language, is constructed in this paper. A concrete algorithm, which can change natural language to the medium language, is programmed, too. The medium language is composed of query link, concept link, query aim link and query condition tree etc.5. After the careful study of the outer character and the inner semantics of the Chinese query statement, the classification of the query statement's sematic templates is given out, and the classifying of algorithmes of the Chinese query statement from form to sematic has been worked out, the classifying algorithm can be implemented by computer. In the past, people only studied the imperative sentence and special question, now the yes-or-no questions and the positive and negative questions are also studied. In the syntactic analysis, the rules of clearing of the pronouns and the complement of the omitting of the query statement have been studied, and the algorithms of the software are realized.6. Through the study of the query aim of the Chinese query statement, the conception of the query aim is decomposed into three phases, there is direct query aim, logic discursion aim and compare judge aim, the relation of the three phases has been studied, and a set of conversion algorithm of each other has also been designed. Through the study of all kinds of query conditions of the Chinese quer...
|
PDF Full Text Request