Different kinds of Knowledge graphs and their use cases
Knowledge graph (KG) has been widely used in different industries, including finance industry, medical industry, online shopping and of course IT giants. Many governments have also seen the advantage of using Knowledge graphs too. And these knowledge graphs differ mainly in their use cases, which determines the design of ontology, Knowledge graph and input datasets. In this blog article, I’ll introduce several representative Knowledge graphs and point the readers to the possible corresponding sources for further reading. These knowledge graphs mainly differ in the data content, data sources and use cases.
General world Knowledge graph
General Knowledge graph refers to the Knowledge graphs contain general knowledge about the world (e.g. Wellington is the capital of New Zealand, Van Gogh died in 1890, etc.). They are often used for search engine, chatbot and personal assistant such as Siri or Alexa. One visualized example of how Google search uses Knowledge graph to enhance the search experience is given in the previous blog article (https://caohongliu.medium.com/knowledge-graph-an-introduction-for-beginners-c7ad82b5082a). By May 2020, the information covered by Google’s Knowledge graph had grown to 500 billion facts on 5 billion entities [1]. Similarly, Microsoft also has a general world Knowledge graph: The Bing knowledge graph, which contains information about the world (such as people, places, things, organizations, locations, and so on) and powers question answering on Bing [2].
Product Knowledge graph
Product knowledge graph is the kind of knowledge graph which contains the semantic knowledge about the products, their properties, the relations among the products and the relations between the products and the general world. Product Knowledge graph can help users to find what they want easily and enhance the user experience. There is no surprise that all the giant e-commence companies like Amazon, Alibaba and eBay all have their own Product Knowledge graphs. For example, in Alibaba, after a user purchases a grill and clicks on charcoals, with their Product Knowledge graph, they will infer a concept that this customer is planning an “outdoor barbecue” (infer the needs for the user and recommend relevant missing items for an outdoor barbecue) and connect this concept with other concepts such as “keep warm for your children” as there’s temperature drop next week. “Outdoor Barbecue” and “Keep Warm for kids”, is introduced as bridging concepts connecting user and items to satisfy some high-level user needs or shopping scenarios [3].
Social Knowledge graph
Social Knowledge graph contains information about people, relations among people and the things people like such as movies, music, celebrities and so on. Facebook is known for having the world’s largest social graph. Facebook engineers have built technology over the past decade to enable rich connections between people [2]. With Social knowledge graph, Facebook is building a deeper understanding of not just people, but also the things that people care about. Unlike Product Knowledge graph which infers user interest from the products they buy/click/save, Social knowledge graph can infer user interests from a broader aspects such as what a user liked/listened/commented/bought/clicked/watched and what her/his friends liked/listened/commented/bought/clicked/watched.
Academic Knowledge graph
Academic Knowledge graph contains information such as people (mostly researchers), publications, fields of study, conferences, journals, affiliations as well as their relations (e,g. citations, co-authors and so on). It allows a user to see connections between researchers and pieces of research that may otherwise be hard to determine [2]. The figure below is an overview of Microsoft Academic graph, which contains 209792741 papers, 253641783 authors, 52431 affiliations by 11/2018. Many researchers also collected the publications of Covid19 to create a Covid Knowledge graph to help the research in this field and generate report automatically (a demo can be found in [4]).
Enterprise Knowledge graph
I hesitated if I should add Enterprise Knowledge graph (EKG) to the list, because I think some company might need different KGs for different use cases while EKG sounds that a company only need this single KG. According to [6], “An enterprise knowledge graph is a representation of an organization’s knowledge domain and artifacts that is understood by both humans and machines. It is a collection of references to your organization’s knowledge assets, content, and data that leverages a data model to describe the people, places, and things and how they are related”. I prefer to call such KGs as Domain specific Knowledge graph, as they provide some specific, deeper knowledge on a domain/field. For example, JPMC’s KG on the finance domain, AstraZeneca’s KG on the medical domain and so on. These EKGs can help the company to make better decisions such as better investment or drug discovery suggestions.
Conclusion
There are also many other kinds of Knowledge graphs not listed here. For example, The LinkedIn Economic graph contains entities such as people, jobs, skills, companies, locations, which can provide economy level insights for countries and regions. However, the objective is not to give an exhaustive list, but to give an overview of some popular KG types and how they differ from each other by data source and use cases. Thinking about use cases is very important when designing a Knowledge graph, because the use cases can help you to decide which data sources to use, what ontology you need, and what questions your KG need to answer. In this blog, I talk about the differences among different kinds of KGs, but one thing remain in common is that all the KGs try to give a holistic view or a god view of their center of interest.
References
[1] https://en.wikipedia.org/wiki/Google_Knowledge_Graph
[2] Noy, N., Gao, Y., Jain, A., Narayanan, A., Patterson, A., & Taylor, J. (2019). Industry-scale knowledge graphs: lessons and challenges. Queue, 17(2), 48–75.
[3] Luo, X., Yang, Y., Zhu, K. Q., Gong, Y., & Yang, K. (2019, November). Conceptualize and infer user needs in e-commerce. In Proceedings of the 28th ACM International Conference on Information and Knowledge Management (pp. 2517–2525).
[4] http://159.89.180.81/demo/covid/Covid-KG_DemoVideo.mp4
