Read about how knowledge graphs address use cases in the financial sector from market intelligence to regulatory reporting, improving the use and reuse of data.
An ontology is a formal and systematic way of representing knowledge within a particular domain, including the concepts and the relationships between them. It involves specifying individual components, such as objects and their attributes, as well as rules and restrictions governing their interactions. This structured representation of knowledge not only allows for more efficient sharing and reuse of information but also facilitates the discovery of new knowledge within the domain.
Ontologies can be applied to collections of facts to create knowledge graphs. Unlike other formal specifications for knowledge representation, such as taxonomies and thesauri, ontologies express complex relationships and enable the linking of multiple concepts in a variety of ways. They are the most sophisticated instruments available to express domain knowledge in human and machine-readable forms.
In this post, we present several of the key benefits they offer and support them with case studies of Ontotext clients and other examples.
In the context of the Financial Services Industry domain, the most popular examples of such data are entity (Who?) and product (What?). These two key data elements are used in approximately 80% of the use cases in the sector.
A Financial Services Industry knowledge graph, powered by an ontology, typically contains domain models and instance data about:
This precise representation helps accurately describe and manage financial data and is essential for analytical insights (Company 360, Market/Industry Insights, Risk and Opportunity Events) or regulatory-driven use cases (KYC, Sanctions, AML).
An ontology can represent the hierarchy of entities that distinguish between private and public corporations, partnerships, governmental and non-governmental organizations, funds, trusts as well as the relationship between them. This is illustrated in the following two instances.
The Financial Industry Business Ontology (FIBO) specifies over 70 different autonomous agents, defined as “something autonomous that can adapt to and interact with its environment”. Subclasses of this class are all persons, organizations and corporations, arranged in a logical hierarchy.
The following diagram shows a part of this graph, centered around legal entities.
Ontologies are used to describe not only classes, but also the relationships between these classes. The diagram below shows the relationships between JPMorgan and The Federal Reserve as a relationship between a regulator and a regulated entity.
In their talk: How Knowledge Graphs and Graph Databases Take Your Data Further, presented at Ontotext’s Knowledge Graph Forum 2022, Ragini Okhandiar and Krishna Potluri from JPMorgan Chase & Co. share their expertise in building an in-house knowledge graph.
They provide practical examples of the requirements an enterprise might have as well as the shortcomings of the traditional warehousing approaches. Their talk also explains the limitations of data catalogs and gives clear examples of the advanced capabilities offered by knowledge graphs. The following slide summarizes the benefits and the unique solution features that knowledge graphs can contribute to the business needs of an enterprise.
Ontologies can also facilitate the data sharing and integrations across various financial systems (banks, investment firms and regulatory agencies), by providing a common vocabulary to describe and exchange data.
FIBO represents such a common vocabulary. It is reused in modeling the publication of entity data or regulatory-mandated data exchange, as seen in the example provided below.
In their Transforming Data Collection action plan, The Bank of England and The Financial Conduct Authority rely on ontological modeling and data representation to achieve key objectives:
Know Your Customer (KYC) and Sanctions are two critical components of the regulatory compliance processes. Here, the ability of knowledge graphs to integrate diverse data from multiple sources is of high relevance. As you can see from the slide below, knowledge graphs can provide a single access point for various types of data such as structured data, knowledge organization systems, transactional data and signals from unstructured content.
This capability can be applied to KYC and Sanctions use cases, where building a profile of a party and integrating all sorts of signals is of great importance. For example, as the slide below shows, it can help quickly deliver a graph-powered system of effective sanctions compliance.
Another advantage of using financial ontologies is that they can be reused across different projects, institutions or applications. This reduces the need to start from scratch when developing new systems or integrating existing ones. Thanks to their modularity, ontologies can be easily modified, expanded or refined to accommodate new concepts, regulations or market developments as the industry evolves.
Refinitiv reuses part of FIBO’s Entities and Corporate bodies ontology when publishing company data in the permid.org related products. You can see more details about Refinitiv’s use of ontologies and FIBO on the permid terms page.
Ontology-based financial systems can also leverage automated reasoning tools to derive new insights or make predictions based on existing knowledge. For example, they can detect hidden relationships between entities, suggest suspicious transactions or identify investment strategies or opportunities.
An ontology-based knowledge graph can infer relationships between different financial entities. For example, it can identify subsidiaries of a parent company or detect hidden ownership structures that may be indicative of reputational risk, fraud or regulatory violations.
As humans, we are constantly using abstractions. Depending on the level of detail, we might use a high-level abstraction or be very specific in our expression. The ontological expression of domain knowledge allows us to map and align the abstractions we use and break them down into their components. In this way, we can achieve normalization of the meaning behind the concepts we use and improve our common understanding.
Let’s have a look at an illustration of how humans use abstractions and how different users might need different levels of granularity. An IT manager might be dealing with software, hardware and data, while an expert might make further distinctions in each category (for example between laptops, servers, mobile devices, etc.). These abstractions will serve both the IT manager and the expert and here the shared ontological model can align these two points of view.
By having a common ontology, which unifies standard industry classifications (such as GICS, SIC, NAICS, NACE) and enrichment by custom classifications of activities, an investment analyst can compare the investment allocations of two companies.
An ontology-powered knowledge graph can improve search and querying capabilities by enabling more accurate and relevant results as well as more expressive and flexible querying based on semantic information.
When we combine the previously described capabilities for detailed knowledge representation, new data coming from inference and alignment of abstractions, this helps create advanced searches. As you can see in the screenshot below, such an advanced search can match the optimal investment target.
Another capability of knowledge graphs that contributes to improved search and discoverability is that they can integrate and index multiple forms of data and associated metadata. As a result, they can provide combinations between multiple search, discovery, ranking and classification capabilities as shown in the diagram below.
Investment banking is one of the most competitive lines of business. Here a transaction advisor can leverage a knowledge graph to gain competitive advantage and information arbitrage.
Another benefit of ontologies is that they provide a structured framework for organizing, maintaining and using financial domain knowledge. This makes it easier to manage and update information as the industry changes.
In the Principles for the Sound Management of Operational Risk, The Bank of International Settlements has recommended that every operational risk can be classified into categories. These categories include internal and external fraud, clients, products, business practices, delivery and others.
A knowledge graph-powered system, utilizing the FIBO-V product, can power a solution that enables the capabilities in the generalized use cases as shown in the diagram below.
Many content publishers have transformed themselves to insights providers, augmenting their portfolio with data and niche analytics products. These strategic initiatives are supported by adopting a sophisticated ontological model, which is the foundation for a future-proof products and offering portfolio
Ontology-powered knowledge graphs have multiple advantages for the Financial Services Industry, such as improved knowledge representation, easier data interoperability, reusability and extensibility of ontologies, automated reasoning and inference, alignment of abstractions, improved search and querying, content knowledge management and more.
All these use cases can benefit greatly from the building blocks of knowledge graph technology as discussed in this blog post and summarized in the diagram below. These blocks work together to create unique foundational capabilities and result in various value drivers.
All in all, ontology-powered knowledge graphs provide a sophisticated and accurate way to represent domain knowledge in human and machine-readable forms. This is essential in facilitating complex financial concepts representation as well as data sharing and integration. We hope that the provided practical examples highlight the importance of using such knowledge graphs to accurately interpret and manage financial data, enable analytical insights and comply with regulatory requirements.