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.
For over two decades, Ontotext has been building knowledge graphs for some of the world’s most knowledge-intensive enterprises – all the way from Life Sciences and Financial Services to Publishing, Government and Industry. Today, the knowledge graph is no longer an exotic thing and has been adopted by enterprises all over the world.
As time goes on and more and more enterprises are turning to knowledge graphs for unlocking the value of their data assets, there are recurring questions we get asked by prospects who want to see what our company’s take is and how our technology can address their challenges.
Here are 5 commonly asked questions we’ve been getting in the last few years:
Peio Popov: I think that our customers choose us because of four main features of our technology.
The first one is related to big graphs and metadata. The graphs we build contain billions of statements, with over 1M entities and concepts. They can link entities across more than 5 data sources and can include 1M documents with over 100 knowledge graph tags per document.
The second essential feature is query performance. Although the ability to ingest data from disparate sources is very important for a knowledge management solution, when you actually start using the data, query performance becomes even more important. What we offer here is 10 transactional updates/sec on master data, 500 updates/sec for documents and metadata, 100 graph queries/sec/node (including inferred facts), RDFS+ reasoning, 1000 full-text searches/sec across docs and data, etc. These figures give a pretty good idea of the foundational requirements we can satisfy for a production system.
The next feature is text and graph analytics. Here, we can efficiently extract new entities and facts from text, retrieve similar documents and entities, automatically perform classification and link prediction, rank relevance and importance, etc.
Finally, our experience shows that enterprises need serious production capabilities and regarding operations and data quality we can offer multi-DC deployment across continents, commodity workers, daily updates from external data sources, ability for metadata and instance data curation, etc. All this, I’m happy to say, has been a core feature of our technology. And the other thing, of course, is data quality because, when you have your data and you are ready to run production systems on top of it, it is crucial to be able to monitor and take care of the data quality.
Peio Popov: Our technology powers several types of use cases. The first group of use cases is in market intelligence, including things like Company Intelligence, M&A Investment, Asset Management, Industry Intelligence, etc.
The second group has to do with regulatory and risk. Here we have Compliance, KYC, AML, Sanctions, Trader Surveillance, etc. This also includes reporting capabilities such as Managerial Reporting, Compliance Reporting (for example, SR14, BCBS239, GDPR, etc.).
And finally we have operational resilience types of use cases like Connected Inventory, Data Catalogs, Data Lineage, etc.
Peio Popov: I would say that there are two distinct paths to success.
The first one is to create a Data Catalog. As we know, every data management methodology prescribes this approach. The Enterprise Data Catalog (EDC) is an inventory of data assets – a corporate resource where data can be found as well as a repository of metadata about data stores, complete with locators to find the data, information about who is responsible for it, and who has access to it.
As a result, you can discover what data is available. You can see where it is stored and how it is accessed as well as what authorizations you need to access it. You get a better understanding of what the data means and what its quality is. It’s easy to collaborate with others – you can see who else is using this data or who can help you with business or technical issues. And finally you can control the consumption interface and format.
Here the basic usage scenarios usually include things like search, exploration by ontology, terms, annotations, etc. You can also identify known systems that process this data or its derivatives. You can check flows and provenance or display metrics such as volumes, changes, quality, etc. So, I would say that this is the basic horizontal way to approach your data catalog with the knowledge graph.
The other path is vertical and it’s to focus on key data assets for your business. Typically, in Financial Services, there are two essential enterprise data assets: entities and financial products. Treating data as the main asset and a product improves the usability and the quality of these data assets. This means you need a smooth data ingestion process, flexibility in expressing your data model, different interfaces for consuming the data in the best possible way for each stakeholder and, of course, ability to ensure resiliency and availability.
The typical usage scenarios on this path are ingesting SQL or tabular data, accessing SQL data virtually, using GraphQL as an API standard, read and write to Kafka streams and topics, integrating with common business analytics tools, accessing endpoints from popular data science tools, etc. This is only a small sample of the general capabilities for various types of data provided via a different type of APIs or content streams, integrated with business analytics tools and so on.
Peio Popov: One thing the use cases in Financial Services have in common is the key data assets and some of these key data assets are entities and securities data. The great thing is that Ontotext has extensive expertise in company data and we can easily bootstrap enterprise entity intelligence systems. This is very important because usually the first thing that comes up when we discuss data with our customers is how to get their data in the knowledge graph.
The challenge here is that enterprises have various types of data. They have reference data, which is the structured data that lives in different Reference Data Systems (e.g., company name, structure, industry code, officers, location, brands, etc.).
Then they have information about the knowledge organization, which includes internal metadata, industry codes (e.g., GICS, NAICS, etc.), compliance and reporting (e.g., Basel II, SR14, IFRS); industry Ontologies such as FIBO and so on.
Then there is also all the data connected to transactions and pricing. This is the actionable, operational data, which is typically found in transactional databases and includes investments, acquisitions, issued instruments, portfolio, market data, etc. These are systems that are the heart of each organization and it’s very valuable to incorporate these datasets in the company knowledge graph.
FInally, they have signals from unstructured content, which is the data that has not yet entered the database. These include M&A events, role changes, corporate actions, etc. All this is information that is yet to influence the assessments in the knowledge graph.
Peio Popov: One essential benefit of our technology is the ability to integrate all types of data into one knowledge graph and make it easy to use. This allows you to discover and identify the key assets of your business and to predict risk and opportunity scenarios. It also enables you to structure, unify and standardize your meaning across the organization as well as to measure quantitative and qualitative information. All this is important because having a holistic view of the enterprise and all its assets facilitates decision making based on deep insights analysis.
Other important benefits are having your data standardized in a way that is compliant with standards and regulations, which reduces risk error, being able to work with your data at scale and in an agile manner, and so on.
As a result of all these improvements, you are doing the same you are doing right now, but you are doing it much better. So, another type of benefit is cost optimization of things like risk, controls, time, repeatable manual efforts, physical and IT infrastructure expenses, etc. And not only that! Our approach also enables new revenue from the addition of new and better monetization of existing information and data assets. This is possible because, when you have your data handy, easily reusable and exposed for different types of usage, new revenue comes easily.
Finally, there are also managerial benefits such as real-time big-picture analysis and drill down of insights, ability to view key success and failure factors, easier assessment of company capabilities, improved ability to analyze performance and ROI as well as to audit and demonstrate ongoing compliance, etc.
So, it is not surprising at all that knowledge graphs are becoming increasingly important and form the backbone of many enterprise knowledge management solutions.