斯坦福课程Knowledge Graphs-What is a Knowledge Graph?

1. Introduction

Knowledge graphs have emerged as a compelling abstraction for organizing world’s structured knowledge over the internet, and a way to integrate information extracted from multiple data sources. Knowledge graphs have also started to play a central role in machine learning as a method to incorporate world knowledge, as a target knowledge representation for extracted knowledge, and for explaining what is learned.

Our goal here is to explain the basic terminology, concepts and usage of knowledge graphs in a simple to understand manner. We do not intend to give here an exhaustive survey of the past and current work on the topic of knowledge graphs.

We will begin by defining knowledge graphs, some applications that have contributed to the recent surge in the popularity of knowledge graphs, and then use of knowledge graphs in machine learning. We will conclude this chapter by summarizing what is new and different about the recent use of knowledge graphs.

2. Knowledge Graph Definition

A knowledge graph is a directed labeled graph in which the labels have well-defined meanings. A directed labeled graph consists of nodes, edges, and labels. Anything can act as a node, for example, people, company, computer, etc. An edge connects a pair of nodes and captures the relationship of interest between them, for example, friendship relationship between two people, customer relationship between a company and person, or a network connection between two computers. The labels capture the meaning of the relationship, for example, the friendship relationship between two people.

More formally, given a set of nodes N, and a set of labels L, a knowledge graph is a subset of the cross product N × L × N. Each member of this set is referred to as a triple, and can be visualized as shown below.

The directed graph representation is used in a variety of ways depending on the needs of an application. A directed graphs such as the one in which the nodes are people, and the edges capture friendship relationship is also known as a data graph. A directed graph in which the nodes are classes of objects (e.g., Book, Textbook, etc.), and the edges capture the subclass relationship, is also known as a taxonomy. In some data models, A is referred to as subject, B is referred to as predicate, and C is referred to as object.

Many interesting computations over graphs can be reduced to navigation. For example, in a friendship knowledge graph, to calculate the friends of a friends of a person A, we can navigate the knowledge graph from A to all nodes B connected to it by a relation labeled as friend, and then recursively to all nodes C connected by the friend relation to each B.

A path in a graph G is a series of nodes (v1, v2,…, vn) where for any i ∈ N with 1 ≤ i < n, there is an edge from vi to vi+1. A simple path is a path with no repeated nodes. A cycle is a path in which the first and the last nodes are the same. Usually, we are interested in only those paths in which the edge label is the same for every pair of nodes. It is possible to define numerous additional properties over the graphs (e.g., connected components, strongly connected components), and provide different ways to traverse the graphs (e.g., shortest path, Hamiltonian path, etc.).

3. Recent Applications of Knowledge Graphs

There are numerous applications of knowledge graphs both in research and industry. Within computer science, there are many uses of a directed graph representation, for example, data flow graphs, binary decision diagrams, state charts, etc. For our discussion here, we have chosen to focus on two concrete applications that have led to recent surge in popularity of knowledge graphs: organizing information over internet and data integration.

3.1 Knowledge Graphs for organizing Knowledge over the Internet

We will explain the use of a knowledge graph over the web by taking the concrete example of Wikidata. Wikidata acts as the central storage for the structured data for Wikipedia. To show the interplay between the two, and to motivate the use of Wikidata knowledge graph, consider the city of Winterthur in Switzerland which has a page in Wikipedia. The Wikipedia page for Winterthur lists its twin towns: two are in Switerzland, one in Czech Republic, and one in Austria. The city of Ontario in California that has a Wikipedia page titled Ontario, California, lists Winterthur as its sister city. The sister city and twin city relationships are identical as well as reciprocal. Thus, if a city A is a sister city of another city B, then B must be a sister city of A. This inference should be automatic, but because this information is stated in English in Wikipedia, it is not easy to detect this discrepancy. In contrast, in the Wikidata representation of Winterthur, there is a relationship called twinned administrative body that lists the city of Ontario. As this relationship is symmetric, the Wikidata page for the city of Ontario automatically includes Winterthur. Thus, when Wikidata knowledge g