In 1976, a computer scientist named John Sowa wrote a groundbreaking paper for the IBM Journal of Research and Development called “Conceptual Graphs for a Data Base Interface.” The paper was the culmination of research he had been doing since 1962 for IBM’s applied mathematics group. In the paper, Sowa introduced the conceptual graph, a radical new way to represent knowledge in computer-based processing. Conceptual graphs represent data as basic, flexible, dynamic concepts that, taken together, can form other, more complex concepts. Conceptual graphs can infer and express relationships between data. This means they allow computers to engage in making meaning as well as present data.
At the time, the conceptual graph was revolutionary, particularly for natural language processing (the branch of computer science that deals with the interaction between computers and human language). Sowa’s conceptual graphs inspired a range of new approaches to using computers to achieve goals and find answers. The conceptual graph was a giant leap forward in the field of artificial intelligence.
For John Sowa, the invention of the conceptual graph had another auspicious consequence beyond its application to computer science. The conceptual graph made Sowa someone others seek out when they want to test their ideas about artificial intelligence. It connected him to many young computer scientists, some of whom, forty years after the invention of the conceptual graph, carry on his legacy.
Today Sowa is a fellow at Kyndi, a pioneering company in natural language processing, where he works alongside his protégé of twenty years, Arun Majumdar, the company’s founder and chief scientist.
Sowa met Majumdar some twenty years ago at a technology conference. In 2001, Majumdar so impressed Sowa with his work on conceptual graphs, Sowa took early retirement from IBM so he could form a company with Majumdar. The company was called VivoMind Intelligence. Scientists at VivoMind developed a way to analyze massive quantities of information using a patented technique called cognitive memory. In one document-analysis project using cognitive memory, the team at VivoMind was able to complete a project in eight weeks that other companies estimated would require forty people two years (that’s eighty person years).
Referring to Watson, the IBM supercomputer that defeated humans on the Jeopardy TV show, Sowa said, “If you had plugged the cognitive memory into the IBM Watson system, you could answer the Jeopardy questions in just logarithmic time, with just a laptop instead of a supercomputer.”
Kyndi was formed in 2014, with John Sowa joining founders Arun Majumdar, Ryan Welsh, and Paul Tarau in order to refine cognitive memory as well as other patented methods of analyzing large amounts of data. These technologies now include approaches to automating domain ontologies, relating multiple documents to one another in analyses, and using symbolic processing methods along with mathematical techniques drawn from physics to pinpoint information.
What started in 1976 as “Conceptual Graphs for a Data Base Interface” continues to evolve. The legacy of John Sowa in artificial intelligence is secure and still growing.