Knowledge Graphs: An Overview
What is the knowledge graph?
In my previous posts, we have talked about graphs, ontologies in short and in detail. In this post, we will see the knowledge graphs from a bird’s view and gain some intuition about them.
Ontology is described as the study of what exists in philosophy whereas described as a specification of the meanings in Artificial Intelligence. That is, an ontology specifies the conceptualizations, describes the types, properties, and interrelationships between entities in an information system. We have been using ontologies in Artificial Intelligence studies for more than forty years but in the past decades, the knowledge graphs have shed some spotlights into the field. The main reason behind their popularity is their usage by the numerous tech giants like Google. But, what even is a knowledge graph?
What is the knowledge graph?
Like taxonomy or ontology, the definition of the knowledge graph is not agreed upon. There was even a paper, Towards a Definition of Knowledge Graphs [1], which noted the several descriptions and helpfully tabulated them.
The above-mentioned paper then goes on to propose the following definition of a knowledge graph:
A knowledge graph acquires and integrates information into an ontology and applies a reasoner to derive new knowledge.
Let’s discuss some common factors mentioned in the table:
1. A knowledge graph is a graph
A knowledge graph is organized as a graph that makes use of the relationships in the data and provides huge flexibility to the user.
2. A knowledge graph is semantic
In knowledge graphs, the meaning of the data comes with the data, in the form of the ontology. That is, data can be expressed in terms of the entity it belongs to or the relations it has with the other entities. Thus, querying is easier.
3. A knowledge graph provides inference
Since knowledge graphs are based on the ontologies, they provide some form of inference, that is, derivation of some implicit information from the explicit data is possible by making use of the various graph-computing techniques.
Ontology vs Knowledge Graphs
Knowledge graphs are one of the ways to represent ontologies which consist of the collection of entities where the types and properties have values declared for them, and where the relationships between them are mapped.
Usually, an ontology deals with the concepts, not the instances of these concepts. For instance, there could be the relation that the Rosé is a wine, but not the fact that the Rosé on the shelf in our local market. If an ontology is represented with a knowledge graph, it is possible to extend this knowledge graph with the fact that it is extracted from some source and could help in the interpretation of the given source.
Google’s Knowledge Graph
The Knowledge Graph is an important part of the search experience in Google. It collects data through various sources such as Wikipedia, CIA World Factbook and Freebase about people, events, animals, events, history, and other topics in order to enhance its search engine’s results.
If you search for a specific person, the Google Knowledge Graph will show an almost complete profile, depending on how well the person is.
- For instance, in the case of Alan Turing, apart from the search results on the left, we can see his profession, birth and death dates, known partners, won prizes, education history, some quotes belong to him, and some related person recommendations on the right as a panel.
Google’s Knowledge Panel/Card shows the most visible result of the work the knowledge graph does behind.
- Searching for a recently released movie, for instance, Joker will show posters, reviews, the show places in your local as well as the showtimes in your current GMT.
When Google released the Knowledge Graph in 2012, they made a really introductory and explanatory video. This explains how exactly the knowledge graph works and how it affects the results you get when you search for a specific term.
All in all…
Most of the time, knowledge comes from multiple sources and must be integrated and knowledge graphs are pretty good at it. Several companies or groups create their own version of the knowledge graph to organize information into data and knowledge such as Knowledge Vault, Microsoft’s Satori, Facebook’s Entities Graph, and Google Knowledge Graph. In this post, we made an overview of knowledge graphs and try to understand the actual meaning of it. For more detail, please stay tuned!
