Despite the proliferation of Knowledge Graphs, there is not one commonly accepted definition of the technology. One of the definitions found on the web describes Knowledge Graphs as a specialization of semantic networks where meaning is expressed as structure, statements are unambiguous, and a limited set of relation types are used. Unambiguity of statements in Knowledge Graphs is further achieved through unambiguity of units the knowledge graph is composed of. This definition was first introduced by van de Riet and Meersman (van de Riet 1992), Stokman and de Vries (Stokman 1988), and Zhang (Zhang 2002)1. It’s worth noting that there are some platforms and system that claim to be knowledge graphs but don’t necessarily fall into the category as it is defined above. Yet, the definition of the Knowledge Graph technology isn’t strict and multiple interpretations are possible.2
Knowledge Graph is not a new technology. With over 20 years of history, graph-based knowledge representation applications have been thoroughly researched and are seen as a part of a larger domain of science pertaining to knowledge representation and reasoning.
Coined by Google in 2012, Knowledge Graph term as we know it today was popularized in the context of Google search, representing Google’s ability to return search results in the form of “things” and concepts rather than strings of information. At the core of Google’s Knowledge Graph are millions of objects found on the web and enhanced with semantic information. Understanding the meaning of each object enables Google to identify relationships between them and then present it in a consolidated format to users as an answer to a search query. Google’s Knowledge Graphs are capable of acquiring new data as it arrives and can find and collect the provenance of data, which makes them highly intelligent applications. The source of information, the information itself, and the way the information was acquired - all this data can be stored in a Knowledge Graph.
In reality, the history of Knowledge Graphs didn’t start from Google. There are three distinct periods that can be defined3:
From the moment the language was invented, people used it to accumulate and pass knowledge. Semantics Networks were one of the first computer-understandable ways to represent knowledge. Created by Collins and Quillian, Semantic Networks theory introduces us to a system consisting of nodes, that represent objects, and edges that represent relations between objects.
Linked Data is a framework that uses RDF to build graph-based data models and OWL to define data annotations. Linked Data builds on the theory of Semantic Networks but addresses some of the Semantic Networks limitations. In particular, Linked Data framework restricts multiple interpretations of constructors and allows users to define the meaning of labels on nodes and edges.
Google introduced the term Knowledge Graph to a wider audience as a technology utilized in search. Google Knowledge Graphs are presented to users in the form of a small box on the left side of search results. The box contains information about the search query plus it displays additional information relating to the query. When searching for Leonardo da Vinci, a user sees not only an excerpt from Wikipedia with biographical information, but also a list of artworks and people who are searched for along with Leonardo da Vinci - Michelangelo, Raphael, and Vincent Van Gogh. By emphasizing relations between searched objects, Google prevents users from having to manually discover multiple links and enables them to gain a comprehensive understanding of the sought object right from the start.
Other technology companies picked up on the trend and came up with their own knowledge graph-based technologies, including Facebook with its Social Graph Search, LinkedIn with Economic Graph, and Microsoft Satori - a knowledge graph technology powering Bing search. Startups centering around knowledge graphs also emerged with Diffbot being the most prominent example. Among university projects YAGO and DBpedia are worth mentioning.
In 2018 Gartner included Knowledge Graphs into its 2018 Emerging Technologies Hype Cycle report, thus proclaiming the technology one of the many that can bring a firm or a startup a competitive advantage.
As the volumes of data grow, utilizing technology that helps organize the data and present it to users in a simple and clear form, empowering learning and knowledge sharing, becomes increasingly important. Implementations of knowledge graphs can be seen in different fields but three are especially apparent:
Knowledge graph applications on the web are often viewed as attempts to get closer to the vision of the Semantic Web. What is the Semantic Web? It’s an idea expressed by Tim Berners-Lee in Scientific American Magazine, presenting the web as a collection of web content including objects, links, and transactions that is meaningful to computers.
Understandably, knowledge graphs on the web are mostly viewed as a tool to transform web from a collection of websites and links to a full-scale knowledge base, where searching the web is equal to reading a book where the information flow is predefined by someone else in order for a reader to gain the best possible understanding of the subject and where an answer to the initial question already includes answers to questions that might follow.
The Semantic Web vision implies that it will be possible to search for any kind of data from documents and links to relationships, background information, contacts, location data and etc. For this idea to become a reality one thing needed to happen - a set of standards had to be invented and used by everyone on the Web. The standards do not affect the content of web pages but inform software programs about the meaning of content including the meaning of additional information unavailable to visitors such as where the content was derived from, who created it, and even where the authors were born and educated.
Several standards emerged following the efforts to build the Semantic Web. Among them are RDF (Resource Description Framework) and OWL (Web Ontology Language). These standards were, however, uneasy to apply across the whole universe of the Web. A more practical vision of the Semantic Web was offered by DBpedia - a large RDF dataset. The project aims to extract structured data from Wikipedia, enabling users to find information across multiple Wiki pages. Another project that further developed the idea of the Semantic Web is schema.org - a collective initiative of major search engines such as Google, Microsoft, Yandex, and Yahoo!. Schema.org provides predefined schemas for the webmasters to mark up web pages. The search engines can understand the markup and deliver better service to the users. In particular, the integration of schema.org item types in web pages may lead to the subsequent formation of a Google Knowledge Graph.
One of the way knowledge graphs can be used in an enterprise setting is to aid collaboration and innovation. In an enterprise, there are thousands of collaborators on projects, many of them are located in different departments or even countries. Paired with the need to release solutions in the shortest time possible, enterprises face a difficult conundrum of organizing and managing information in a way that promotes rapid sharing of knowledge and ideas. Knowledge Graphs make the body of knowledge collected in an enterprise available to all the participants. They also allow for adding new information, that is then structured in a way that is easily searchable and understandable to everyone involved. Thus Knowledge Graphs demonstrate the potential to make collective innovation easier and more effective.
It’s obvious that Knowledge Graphs are mainly used in the context of knowledge sharing. Understanding of the meaning of information equips knowledge graphs with the ability to help people master the subject faster and more efficiently. By pulling up relevant data from various sources, relating to topics, ideas, people, projects, products, claims and etc. Knowledge Graphs alleviate the pains of search, reveal related information otherwise undiscoverable to users, and can speed up the pace of learning and improve productivity. In an enterprise, the application of knowledge graphs can also be seen as a knowledge retention measure meant to diminish consequences of workforce aging.
The increase in the volume of data affects not only the data found within the corporation but also the market data, such as information about competitors, audience, and industry trends. Extracting and structuring such information may bring competitive benefit to companies, as relying on internal sources of innovation only is often not enough in a cut-throat market environment. It takes substantial financial and time resources to innovate from within. At the same time, having access to a constantly updated source information in the form of a Knowledge Graph may lead to unexpected insights thus speeding up the process of innovation. Additionally, applications of Knowledge Graphs can be found in sales and marketing in cases when Knowledge Graphs are used to pull up historic and background information on acquired leads.
Community-driven Knowledge Graph applications leverage the shared effort of people to enrich Knowledge Graphs with new information and annotate existing information. These efforts result in the creation of a collectively edited knowledge base. A most notable example of a community-driven Knowledge Graph application is Freebase - a now discontinued collection of structured data consisting of data contributions submitted by registered users. In fact, Metaweb, a company behind Freebase, was acquired by Google in 2010 with an obvious purpose to use Freebase as a foundation for Google Knowledge Graph. Wikidata is another example of a project that uses structured data that can be edited by anyone.
These are just several examples of Knowledge Graph application. Like any technology, the potential of Knowledge Graphs is yet to be fully realized, and other applications may transpire in future.
If you’re looking for ways to integrate Knowledge Graphs into your enterprise, connect with one of our experts on Knowledge-based Applications.
1. RP van de Riet, RA Meersman. Knowledge Graphs. 97 In Linguistic Instruments in Knowledge Engineering: Proceedings of the 1991 Workshop on Linguistic Instruments in Knowledge Engineering, Tilburg, the Netherlands, 17-18 January 1991. (1992). Frans N. Stokman, Pieter H. de Vries. Structuring Knowledge in a Graph. 186–206 In Human-Computer Interaction. Springer Science + Business Media, 1988. Lei Zhang. Knowledge graph theory and structural parsing. Twente University Press, 2002.
2. Ehrlinger, Lisa & Wöß, Wolfram. (2016). Towards a Definition of Knowledge Graphs.
3. Gomez-Perez J.M., Pan J.Z., Vetere G., Wu H. (2017) Enterprise Knowledge Graph: An Introduction. In: Pan J., Vetere G., Gomez-Perez J., Wu H. (eds) Exploiting Linked Data and Knowledge Graphs in Large Organisations. Springer, Cham