用于考古数字博物馆的受限自然语言查询系统

【作者】 马晓娜；

【导师】 杨承磊；

【作者基本信息】 山东大学 ， 计算机应用技术， 2005， 硕士

【摘要】 汉语自然语言查询系统是自然语言理解、数据库技术、人工智能、人机界面相结合的产物。它使用户可以直接以汉语自然语言的方式,向数据库系统发问并获得所需的信息,从而大大改善了人机交互的容易程度。近年来,它作为人机交互中智能接口的重要部分受到重视,成为具有重要理论意义与广泛实用价值的研究领域。 本论文主要的研究目标是结合考古数字博物馆特定的应用需求,研究与实现一个基于受限的汉语自然语言查询系统。它允许用户在界面层输入受限汉语查询语句,然后通过一定的算法把用户的输入转换成标准的SQL语句,并在文物数据库中完成所要求的查询,输出查询结果。 本论文的工作与贡献主要有: 1.提出符合汉语语法并能满足查询要求的受限文法规则。本论文首先通过对山东建筑工程学院计算机系2001级的24名学生进行问卷调查得到汉语查询语句的四种句型,分别为:祈使句、疑问句、省略句和多语句。且分析得到祈使句是使用频率最高的句型(约占70.2%),并据此提出本系统所采用的符合汉语言习惯并能满足考古数字博物馆查询要求的受限文法规则。 2.研究设计了适合考古数字博物馆的系统词典。如何建立一个简单适用的系统词典是自然语言查询系统研究工作的基础和难点之一。本文通过对考古数字博物馆的数据进行了详细地分析,提出并设计了三种系统词典:通用词典、专用词典和关联词典。这三种系统词典在查询语句的分析中分别起到了不同的作用:通用词典可以支持进行汉语查询语句的分词操作,并得到切分后的各单词的词性序列,从而可以利用词性序列和受限文法规则得到查询语句的查询目标和查询条件;专用词典可以用于得到汉语查询语句的标准化形式,为SQL语句的形成进行预处理操作;关联词典可以用于得到各表之间的关联路径,从而解决了汉语查询语句中比较复杂的表间相互联系的表述问题。本文采用SQL SERVER数据库表的形式对它们进行定义。这样既可以和查询中用到的后台数据的定义形式统一起来以便于系统的实现,又可以提高系统对查询语句的分析速度,从而提高了系统的性能。 3.提出了一种改进的分词算法——WSDS(Word Segmentation on Database Semantic)算法。该算法参照和单词相关的系统词典中给出的数据库语义,应用考古数字博物馆的领域知识,利用信息论和运筹学的相关知识消除了切分的歧义更多还原

【Abstract】 Chinese natural language query system(CNLQS) is the result of many subjects, such as the natural language comprehension technology, the database technology, AI, the man-machine interface, etc. With the CNLQS, users can use the Chinese natural language to put forward the question and get the answer from the database system directly. In this way, it can make the interaction of the man and the machine more easily. Recently, CNLQS, which is a very important part of the man-machine interaction, begins to be paid more attention and be regarded as a research field that has the theoretical and the practical value.This paper achieves a restrictive Chinese natural language query system based on the archaeological digital museum application. This system permits users to input a restrictive Chinese query sentence through the user interface. Then it can turn the input into the standard SQL through an algorithm and get the result of the query in a cultural relic’s database.Based on the above background, we have done the following work:1. This paper puts forward the restrictive grammars and rules that are consistent with the Chinese grammars and can satisfy the demands of the query. This system firstly gives four types of Chinese query sentences through the investigation paper, which includes the imperative sentence, the question, the elliptical sentence and the multiple sentence. The investigated objects are 24 students of the Computer Department of Shandong Architecture Institute. Through the analysis, we conclude that the imperative sentence’s using frequency is the highest, which is about 70.2 percent. Based on the above research, we put forward a kind of restrictive grammars and rules, which are consistent with the Chinese language habits and can satisfy the query request of the archaeological digital museum.2. This paper designs a system dictionary that applies in the archaeological digital museum. How to build a simple and appropriate system dictionary is one of the foundation and difficulty of our research work. This paper introduces three kinds of system dictionary, which are general dictionary, special dictionary and associated dictionary by the analysis of data based on the archaeological digital museum. The threekinds of system dictionary have the different function. The general dictionary can be used to resolve the Chinese word’s segmentation question and get the sequence of the part of speech. So we can get the query sentence’s target and condition by using the sequence of the part of speech and the restrictive grammars and rules. The special dictionary can get the standardization form of the Chinese query sentence and finish the pretreatment operation. The associated dictionary can resolve the complicated relation of the tables by the associated path. We define the dictionary in the form of SQL SERVER database. In this way, we can unify the definition form of the dictionary and the data, which can improve the analytic speeds of the query sentences and then increase performance of the system.3. This paper puts forward a new word segmentation algorithm—WSDS(Word Segmentation on Database Semantic) algorithm. With the domain knowledge of the archaeological digital museum, using the information theory and the operational research and referring to the database semanteme in the system dictionary, the WSDS algorithm can resolve effectually the problem of the different meanings of Chinese syncopation and receive the correct syncopating form of the query sentence. We achieve the sentence array and the character string of the types of the sentence based on the WSDS algorithm. It supports the need of semantic analysis and the conversion from the natural language query sentence to SQL. We get the approving conclusion at the end.4. Based on the syncopation of the words, this paper puts forward a similar SQL sentences building algorithm—COS(Condition-Object Segmentation) algorithm. This algorithm achieves the apprehension of the restricted query sentences and the building of the similar SQL sentences.5. This paper can turn the similar SQL into the standard SQL by using the associated dictionary and the method of the associated path search at the end.The main innovations of this paper are as follows:1. This paper puts forward the restrictive grammars and rules that are consistent with the Chinese grammars and can satisfy the demands of the query.2. This paper studys and designs a system dictionary that applies in the archaeological digital museum.3. This paper puts forward a new word segmentation algorithm—WSDS algorithm.4. This paper puts forward a similar SQL sentences building algorithm—COS algorithm.This paper is simply researching an archetype that only deals with the Chinese natural language input by the people. In order to perfect this system and make it into practicality, there are many works that we can do in the field of archaeological digital museum actually, such as: how to improve the feasibility of the system transplant, how to resolve the different meanings of Chinese syncopation ulteriorly.更多还原