Cognaize Expands IDP Platform with Knowledge Graph to Fuse AI and Human Expertise

(Jirsak/Shutterstock)

Intelligent document processing company Cognaize has announced a significant expansion to its technology platform. These new capabilities extend the accuracy of the company’s Financial Insights Engine through a proprietary knowledge graph that Cognaize says provides deeper insights and enhanced automation.

Intelligent document processing is a technology that automates the extraction, processing, and analysis of data from unstructured and semi-structured documents. The technology combines elements of natural language processing, machine learning, and artificial intelligence to process, understand, and classify information within documents, which can improve efficiency and accelerate decision-making.

Cognaize says its platform is seeing increased adoption by financial services organizations globally. According to Grand View Research, the IDP market is anticipated to reach $11.6 billion by 2030.

The Cognaize IDP solution utilizes a combination of AI and humans, or what it calls hybrid intelligence. The company says the platform combines sophisticated financial models that have been trained on over 1.3 financial million documents, including loan applications, SEC filings, ESG-related documents, presentations, or trustee reports, with human “experts in the loop” throughout the document automation process.

Cognaize developed its new capabilities in response to its customers’ needs and requests. Knowledge graphs are conceptual models of the world that map entities, their properties, and their relationships. The Cognaize Knowledge Graph, trained on various AI models for the graphical processing of information, constructs an extensive knowledge network that is verified by human specialists. The company says its Knowledge Graph elevates new document ingestion automation to unprecedented levels.

“At Cognaize, we are building a Knowledge Graph across documents,” Cognaize CTO Vahe Andonians told Datanami in an email interview. He gives the example of processing a bond indenture, where information such as name, issuer, guarantor, interest rate, principal amount, and current outstanding amount is spread across many different documents. “The information on all these features is spread across different documents and it is hard to understand from one document much about the bond itself,” he said.

This graphic shows the hybrid intelligence between AI and humans in the platform’s validation loop. (Source: Cognaize)

The Cognaize Knowledge Graph enables the automatic extraction of additional information about the bond, such as the outstanding balance, from 10-K filings (the annual reports filed by publicly traded companies to the SEC). While the bond indenture names the guarantor, it does not detail any outstanding balances, so if a financial analyst asks, “How much is the Guarantor at risk?” the Knowledge Graph has formed connections between these various documents to deliver an answer to this query.

The company says this technology empowers users to ask questions and receive near-instantaneous answers through the ad-hoc execution of AI models directly within a browser. Additional features noted by the company include ingestion support for the XBRL format, encompassing SEC filings, as well as an AI-powered method for optimally distributing validation tasks to human experts.

Human validation is a key factor of the Cognaize platform. Andonians explains that “The combination of AI and humans plays a crucial role in enhancing the effectiveness and reliability of outcomes by harnessing the complementary strengths of both parties.” AI models efficiently process large amounts of information to reveal patterns and make predictions, while humans possess domain knowledge, critical thinking, and creativity, Andonians says, noting these traits are essential for ensuring AI-generated insights align with real-world contexts.

Cognaize CTO Vahe Andonians

The Cognaize Knowledge Graph is continuously validated through its financial expert users and automatically creates training data through this interaction. “As a result, we are not distinguishing between ‘training’ and ‘deployment’ but see it as a never-ending loop which can be fully automated, or can have data scientists in the middle, as per the decision of the client,” Andonians explains.

Andonians says the Knowledge Graph is a step forward in replicating human intellect through its construction of an intricate model of the world, something he believes will provide a path to true artificial intelligence.

“We will continue to match our rapid pace of technology innovation with our hybrid intelligence approach that puts our customer’s financial experts at the center throughout the document automation process to ensure optimal accuracy and efficiency for the world’s leading innovators in banking and insurance,” he said.

Related Items:

Harnessing Hybrid Intelligence: Balancing AI Models and Human Expertise for Optimal Performance

A New Era of Natural Language Search Emerges for the Enterprise

Overcoming Data Integration Challenges to Unleash Massive Potential: Knowledge Graphs 101