Overcoming information overload with knowledge graphs [Q&A]

Graph

Businesses are generating ever larger quantities of data, much of it in unstructured form. Extracting value from this massive amount of information can be difficult, which is why it can sometimes feel as if there is simply too much data.

Augmented intelligence specialist Yewno believes it has the technology to help people research and to understand the world in a more natural manner, inspired by the way humans process information from multiple sensorial channels. We spoke to the company's CEO, Ruggero Gramatica, to find out more.

BN: What's the difference between data and knowledge and how can you get one from the other?

RG: Information does not equal knowledge. Information is merely the data we feed our brain, and as with a computer, data must be processed if we want to gain insights. Our brains take in information, as a computer takes in data, and transform it into knowledge.

To explain this cognitive process, we can imagine a process that concatenates concepts, derives a context, and draws inferences. This schema essentially illustrates the process our brain undergoes to transform data (or information) into knowledge.

Conceptual representation and semantic context are part of the cognitive process that leads to knowledge. In order to gain knowledge, we need an inferential process, i.e. a mechanism that is the building block of reasoning. An inference can be roughly defined as a logical conscious or unconscious concatenation of concepts, in a context, where the link between any two of those carry the informative element correlating the two concepts.

We can access a large chunk of information, but that does not make us any more knowledgeable of a specific subject matter. In fact, information must be read, interpreted and cognitively understood in order for us to act upon it.

  • First we have Data, unstructured, uncritical
  • Once Data is aggregated we have Information, where analysis can be performed
  • Applying cognitive processes to information -- like inference -- leads to Knowledge
  • A comprehensive status of knowledge leads to Wisdom

BN: Tell us a bit more about knowledge graphs.

RG: Used properly, a knowledge graph provides a framework to extract inferential pathways connecting several objects -- taking points of information and turning them into usable knowledge.

A knowledge graph is a mathematical framework for dynamically gathering and semantically processing a large base of structured and unstructured data and information representing multiple subject areas. By mining the source information, the platform creates an efficient algorithm-based representation of the overall knowledge it provides, allowing for capturing significant relationships -- often hidden -- semantically and quantitatively across complex data sets.

Instead of basing results on individual elements, like a typical database query, it uses algorithms to construct relationships -- the graph or network -- among concepts that are not evident. It shows the relationships between concepts and their connections, derives inferences, and suggests promising areas of intelligence.

As a result, a knowledge graph constructed from a large number of diverse information sources and data sets can be used to make predictions, support decision-making efforts, and identify strategies to find meaningful trends hidden within the information.

A full-fledged knowledge graph can find correlations across essentially any type of data set, doing the kind of inferential work it might take a full team of analysts with access to large information data sets many hours to perform.

BN: What's behind the technology?

RG: Knowledge graphs require a flexible structure; a set of dynamic layers that defines the set of types, properties, and relationships continuously adjusting as new information comes through. While other graphs map a picture of predefined relationships between entities, those tend to be limited, as they can’t easily account for the introduction of 'inferred' data from other emerging relationships.

Imagine, for example, the relationships that can be dynamically extracted when tracking business interactions between companies. If you have access to certain private databases, you can have a supply-chain link between the two companies. But such a relationship might not be enough; it’s important to understand on top of all the factual types of relationships the two entities have in common (i.e. sector, investor relationships, size/market commonalities, affiliation, partnerships etc), that others can be inferred by second- or third-order connections.

For example, if both entities are exposed to political, financial or geographic risks, a link with such emerging relationships will enrich other, already existing ones. Moreover, there are relationships that form as emerging property of the graph itself -- these are defined and emerging inferences.

In other words, the amount of information that gravitates directly or indirectly around any concept mapped into the graph, generates an induced, ontological, ever-changing layer that dramatically enriches the relationships between any entities of the graph.

Knowledge graphs are also actual graphs -- based on mathematical models -- which makes them easily expandable and allows for the application of mathematical graph-theory techniques like network analysis. That ability to process and interpret plain language is at the cornerstone of what makes them unique.

BN: Which industries can this be applied in?

RG: Ultimately, knowledge graphs are a framework to formalize and standardize what has traditionally been a uniquely human problem: the transformation of unstructured and diverse types of information into knowledge; a framework that can be sliced and diced, unfolded even over hundreds of millions of individual objects and billions of relationships.

The intrinsic nature of dynamic knowledge graphs -- projecting within their structure billions of relationships across hundreds of millions of data points -- augments the capability of investment analysts and researchers by screening more information and finding unintuitive inferences that result in faster and better quality of insights. And it will apply to multiple fields of human knowledge, from pharmacology and finance to geo-politics and beyond.

Additionally, many institutional investors today are seeking more than simple profit. They are looking to make a social and environmental difference, stewarding their financial resources toward investments that fulfill their purpose and mission while increasing financial returns.

The field of environment, social, and governance (ESG) investing has grown into a full-blown industry, with wealth managers adding ESG advisors to their ranks and companies reporting on their ESG policies. Institutional investors, hedge funds, and family offices that take ESG considerations into account are seeking new ratings and research methods to guide their decision making. ESG themes can include considerations like climate change, pollution, health and safety, labor standards, human rights, anti-corruption, etc. To adequately incorporate these considerations into a portfolio, smart investors determine their own ESG thesis and analyze a wide variety of information and data sources for investment fit and profit opportunities.

The intrinsic nature of dynamic knowledge graphs -- projecting within their structure billions of relationships across hundreds of millions of data points -- augments the capability of investment analysts and researchers by screening more information and finding unintuitive inferences that result in faster and better quality of insights.

BN: How does it benefit the consumer?

RG: Tools and infrastructure for graph-based knowledge representation are coming to market to help professionals in multiple fields analyze massive data sets in order to find non-obvious connections that can lead to better decisions, providing the ability to find hidden relationships that can lead to profitable discoveries while increasing researcher and analyst productivity.

Knowledge graphs are the perfect tool for providing augmented intelligence. In finance, machine learning robo-advisor platforms provide black-box trading advice, but fail to display the rationale of the output of the algorithm. Knowledge graph-based solutions synthesize a constant flux of company, industry, financial, economic and political information to aid investment professionals in making purposeful and profitable decisions.

Photo Credit:  suphakit73/Shutterstock

© 1998-2019 BetaNews, Inc. All Rights Reserved. Privacy Policy - Cookie Policy.