Google, Facebook and Amazon have been using them for a long time: Knowledge Graphs. Those who organize their data this way can filter out much more knowledge from data and lay the foundation for the use of advanced artificial intelligence. In the meantime, Enterprise Knowledge Graphs are also spreading in medium and large companies.
We explain how the technology works and how it is used in data-driven companies.
Definition: What is a Knowledge Graph?
A Knowledge Graph is a term used to describe NoSQL databases that are organized in such a way that they represent data in their semantic contexts. Objects are represented as nodes, their relationship via edges. Structured and unstructured data can be organized into topic clusters (ontologies) this way. Such a data network changes in structure depending on what new data and information is added.
Knowledge graphs can replace the traditional data organization in relational databases. In these, the data attributes to be recorded in tables are defined in advance. Tables are linked to each other, but the linkage is relatively stiff. And unlike the Knowledge Graph, only structured data can be processed.
Knowledge Graphs are much better suited than relational databases for the modern requirements of the corporate world to evaluate data across departments and systems, to understand it in context, and to show the user hidden connections.
An example of Knowledge Graph
Google introduced the Knowledge Graph back in 2012. When a user types "How do I learn to program?" into the search box, the search engine searches the Internet for more than just the keywords learning and programming. The Knowledge Graph has ontologies related to learning and programming. Therefore, the search engine "understands" the target of the question on a semantic level and can provide higher quality answers.
Knowledge Graph and Artificial Intelligence
A Knowledge Graph is a solid basis for the use of machine learning and advanced artificial intelligence. Data is not stored in isolated systems, but is stored centrally - and also put into meaningful relationships with one another. AI tools evaluate data with the help of knowledge graphs in the same way as human brains work. They take into account the meaning and context of individual data.
The use of this technology is not yet established in companies. But increasing data volumes and high competitive pressure are forcing companies to organize their data in a more value-adding way. The proliferation of personalized product and action recommendations, chat bots and navigation aids that use Knowledge Graphs is also creating greater awareness of the technology's potential.
According to Gartner's AI Hype Cycle, Knowledge Graph is one of the AI innovations that will take another five to 10 years before it becomes mainstream technology. For companies looking to achieve competitive advantage, now is the optimal time to adopt the technology and become familiar with its potential.
In Corporate Use: The Enterprise Knowledge Graph
When Knowledge Graphs are used in companies, they are called Enterprise Knowledge Graph. They are not always superior to the traditional organization in relational databases. It depends on the use case.
Let's say we want to organize projects. If budget, staff and timeline are fixed in advance, we are dealing with structured data and a data schema that will not change during the project. We also know in advance which analyses are possible and which should be performed. Provided that we always want to execute projects according to the same schema and always want to evaluate them from the same points of view, relational databases are a well-suited form of data organization.
However, Knowledge Graphs are more suitable if we want to organize many different projects and there are variants and exceptions in the structure of the projects. Does a project need to take grants into account in the financing? Does a project need to be split into sub-projects? Then a Knowledge Graph is superior to the table-based data organization.
It can also better handle unplanned questions for which projects should be analyzed. The Knowledge Graph can group projects by customer, topic, or any other variable and perform cross-cutting exploratory data analysis. IT expertise is no longer required. Any business user is capable of generating such insights.
Use Cases for Enterprise Knowledge Graph
All companies benefit from semantic data organization. Because as soon as knowledge work is required, a knowledge graph can support departments and customers better than traditional database queries.
How does a change in the project affect long-term planning? If an engineering team gives new specifications for a component during a project, this can lead to the supplier having to purchase new machines to implement the requirements. Delivery delays and budget overruns can result. The more complex a project, the more difficult it is to keep track of all the effects. A Knowledge Graph shows obvious and indirect consequences in advance and facilitates risk assessment.
Companies must comply with a wide range of legal requirements and regularly adapt to new legislation. Those who maintain business relationships in several countries can easily lose track of which regulations apply where and when. For example, what tests must product A meet in country B before approval is possible? Is an export license required for delivery in country C? Such questions can be easily queried with the help of a Knowledge Graph - without lengthy manual searches.
Technicians in the manufacturing industry need expert knowledge to inspect, maintain and repair plant and machinery. This is often personal or stored in separate systems. What special tool is needed to repair the machine? Which possible cause of failure should always be checked for the machine? A knowledge graph in maintenance brings together structured data and unstructured empirical knowledge. Causes of faults can thus be identified more quickly, expert knowledge is accessible throughout the company, and customers benefit from shorter technician deployment times.
Product and application data
Which similar products could offer added value for the customer? When selling complex technical products in many variants, statistical models à la "other customers also bought" do not work. However, a Knowledge Graph recognizes which other products in the application environment offer added value for the customer without a large basis for comparison. Because of its semantic approach, it makes the industry and overview knowledge of companies accessible. If a customer specifies specific performance and compatibility requirements, a knowledge graph is able to provide support during the purchasing process and recommend suitable products. In this way, companies increase their closing, cross-sell and upsell rates.
Especially in e-commerce, but also in the industrial supply chain, many companies make it a prerequisite for collaboration that their partners provide data in certain data exchange formats. Without a knowledge graph, organizations have to go to great lengths to do this and classify numerous data manually. Redundancies often arise and the efficiency of data management suffers. With a knowledge graph, data can be stored centrally and output in different formats without redundant data maintenance.
Frequently Asked Questions About Knowledge Graph
The Knowledge Graph stores data in its context and takes semantic relationships into account. It can handle structured and unstructured data. Relational databases can only work with structured data, which they store in tables. These can be linked to each other, but cannot be extended with new variables as easily as a Knowledge Graph. The relational database model is less suitable for AI applications and processing complex relationships.
Data organization as a knowledge graph is suitable both as input for machine learning algorithms and for representing their output. ML algorithms can use data networks instead of relational databases as input for their analyses and improve their results independently. Regardless of the data source, they can create Knowledge Graphs from their results or extend them.
Before the technical development, there is the strategic consideration: For which use case should the knowledge graph be used? Based on this, the relevant data sources are brought together in a NoSQL graph database and semantically organized there. The performance of the Knowledge Graph improves during use. This is because it interprets the activities of the users as feedback, which it uses to better link and weight the data objects.
Better data. Better decisions.