这篇笔记来自于北大邹磊教授的知识图谱讲座（第三讲）

RDF图数据管理（RDF Data Management）

视频地址：https://www.bilibili.com/video/BV1yg4y1v7aD

主要内容：

• RDF Database Systems
  • Background
  • Existing Solutions
  • Graph-Based Approaches
• Distributed System
  • Existing Solutions
  • gStore: a distributed SPARQL query engine

RDF and Semantic Web

• RDF(Resource Description Framework) is a language for the conceptual modeling of information about web resources
• A building block of semantic web（语义网的基石）
  • Facilitates exchange of informatioin
  • Search engines can retrieve more relevant information
  • Facilitates data integration (mashes)
• Machine understandable（机器可理解）
  • Understand the information on the web and the interrelationships among them

1. RDF Database Systems

RDF Introduction

• Everything is an uniquely named resource
• Namespaces can be used to scope the names
• Properties of resources can be defined
• Relationships with other resources can be defined
• Resources can be contributed by different people / groups and can be located anywhere in the web
  • Integrated web "database"

RDF Data Model (RDF 数据模型)

RDF Data Model

RDF本质上一个三元组的集合，可以用图来表示
RDF Graph

RDF Query Model (RDF 查询模型)

RDF Query Model

传统存储查询方式：

• 用关系型数据库来存储三元组数据
• 将SPARQL语句转换为Sql语句进行查询
• 这种方式的缺点是，当数据量大的时候，查询效率非常低。
• 已有的优化方法：
  • Property Tables (属性表)
  • Binary Tables
  • Exhaustive Indexing（完全索引）

Graph-Based Approaches （基于图的方法）

gStore

• We work directly on the RDF graph and the SPARQL graph（利用图的方式来回答SPARQL的查询）
  • Answering SPARQL query = subgraph matching
  • Subgraph matching is computationally expensive
• Use a signature-based encoding of each entity and classvertex to speed up matching
• Filter-and-evaluate
  • Use a false posititve algorithm to prune nodes and obtain a set of candidates; then do more detailed evaluation on those
• We develop an Index over the data signature graph for efficient pruning
  核心思想：用子图匹配的方式来回答SPARSQL查询

Subgraph Isomorphism

• Edge-Join Strategy
• Verte-Join Strategy

Distributed System

RDF Data Volumes

• RDF Data Volumes are growing and fast
  • Linked data cloud currently consists of 325 datasets with > 25B triples
  • Size almost doubling every year
• Linking Open Data cloud diagram

Large RDF Datasets

• Now, RDF datasets become larger and larger
• Yago, Freebase, DBpedia
• Massive volumes of RDF data are growing beyond the capacity of RDF database systems operating on a single machine.

gStore: a distributed SPARQL query engine

System Framework: Overview