關聯式資料庫之關鍵字查詢處理技術

Abstract

當今在網際網路中最普遍的搜尋資訊方式，是使用網路搜尋引擎，給予關鍵字進行查詢。我們希望讓使用者能以接近操作搜尋引擎的查詢方式，給定關鍵字和條件式對於關聯式資料庫內容進行查詢，增加資料庫使用的普及和可用性。本論文之研究目的是對關聯式資料庫提出關鍵字查詢處理技術，能將使用者查詢自動轉換成一個對應的結構化查詢語言的SQL語句，並能符合使用者的搜尋意圖。關鍵字查詢處理技術是將查詢關鍵字進行拆解，拆解成一至多個查詢條件並且分組。接著透過給定設計查詢條件單位以邏輯運算子組合運算的優先順序指定括弧位置，產生最可能表達使用者給予關鍵字查詢的搜尋意圖的條件組合。在決定了查詢條件和邏輯順序組合後，接著轉換成所對應的SQL查詢語句，並且執行於關聯式資料庫。此外，本系統還會依統計觀點分析查詢結果中欄位值的一致性和分佈變化差異進行結果摘要。實驗結果顯示本研究方法之關鍵字查詢處理技術在多數情況下能正確處理單一屬性值、資料型態和屬性值組合、數值條件式的查詢。而所提出的摘要方法也有助於使用者得知隱含在查詢結果的重要資訊。By giving keywords to a search engine on Internet is the most popular way of searching information. In order to improve the popularity and availability of relational database, we would like to provide that users can give a query consisting of only keywords and predicates for database to find the required data from the database, which is similar to the way of giving queries on search engines. Accordingly, in this thesis, we propose the query processing techniques for keyword search in relational databases, which can automatically transform a user's keyword query into the corresponding Structured Query Language (SQL) statement according to the query intent of the user implied by the keyword query. The proposed query processing techniques firstly separate the query into multiple predicate units and perform grouping on these units. By defining the priority of query predicates and the combination of logical operations, the parentheses are assigned to the predicates to generate the combination of the predicates which most possibly represents the query intent of the given keyword query. After generating the combination of the predicates, the query is then transformed into the corresponding SQL query and performed on a relational database. Besides, the system can statistically analysis the query results according to the consistency and the distribution of differences on attribute values to provide a summary of the results. The experimental results show that the proposed query processing techniques for keyword search can correctly deal with queries consisting of attribute value, combined with keywords of data types and conditional value expression in most cases. Besides, the provided result summary can help users find the important information which is implicit in the query results.