基於語義數據存取的Restful/JSON至SPARQL的自動化轉換

Abstract

本研究的目的是建立一個基於句法和語義的線上轉換服務，可以將RESTful /JSON訊息轉換成基於目標語義數據本體設計的SPARQL命令。此服務使得Web客戶端可以直接透過RESTful的方式來對基於本體的RDF資料庫進行操作，且客戶端不必預先知道相關的語義數據本體資訊，也不需要預先送出模式匹配SPARQL命令。

對於每一個數據的操作，本服務會透過最小通過遍歷語義數據的本體結構命令，特別是針對操作對象的屬性和類型，產生一組SPARQL命令。本服務支援三種基本操作：（1）搜索一項類別中的特定實例，（2）檢索一項類別中的特定實例的資料屬性，以及（3）新增一項類別中的新實例，同時透過本體中的基數限制，來保持實例之間的語義一致性。

本服務的主要操作為搜索功能，基於客戶端的輸入資料集來產生SPARQL查詢：原點（查詢限制）集合和目標（查詢投影）集合。這個功能的基本解決方法可以應用在簡單到複雜的情景中。在本研究中，我們針對以下兩種情境的搜索功能進行相關討論：直接推論和間接推論。在直接推論的情況下，搜索功能可以成功解決直接從原點集合中找到的子集之間的交集目標的可達性：在遍歷上的單個跳點。在間接推論的情況下，搜索功能成功解決了直接從原點集合中找到的子集之間的交集目標的可達性：在遍歷上的多個跳點。對於這一點，本研究所提出的方法可以找出兩個類別之間的最短路徑。

This thesis describes the design and implementation of an on-line syntactic and semantic translation service that converts RESTful/JSON messages into SPARQL commands based on the ontology of target semantic data. Such a translation service enables RESTful web service clients to access ontology-based RDF repositories without the knowledge of the semantic data ontology nor the need to issue pattern-matching SPARQL commands. For each data access, the translation service produces a minimal set of SPARQL commands by traversing the ontological structure of the semantic data, especially their Object Properties and Datatype Properties. Three basic operations are supported: (1) searching specific instances of a class, (2) retrieving the Datatype Properties of a specific instance, and, (3) inserting a new instance of a class while enforcing the cardinality restriction specified in the ontology in order to maintain semantic consistency among the instances. The main operation of the translator is the search function, which generates a SPARQL query based on the client’s input data set: origin (query restriction) and target (query projection) sets. The basic solution approach of this function is presented in this document from simple to complex scenarios. Two elemental search cases are discussed: search with direct inferences and with indirect inferences. In the case of direct inferences, the search function solves the reachability of the target set through the intersection among the subsets that are found directly from the origin set: with a single hop on the traversal. For the indirect inference case, the search function solve the reachability from the origin set to the target set: with multiple hops on traversal. For this, the proposed approach finds the shortest paths between two classes. Finally, the current implementation of the translator is described in a detailed manner.