Smart Data Plus Deep Reasoning Equals Business Value from Data Analysis

By on

jz_deepreason_041116Coherent Knowledge would like businesses that want to make more effective use of the knowledge they capture and distribute to become more familiar with Rulelog. That is the logic underlying knowledge representation languages such as Vulcan’s Silk and the rule language used by its own Ergo Suite semantic rules and reasoning platform. Smart Data is no longer just a future dream for many enterprises; it’s now a reality with such systems.

In draft as an industry standard, the knowledge representation and reasoning logic generalizes and complements graph and relational database, semantic, and business rules technologies by enabling flexible representation and deep reasoning with highly complex domain knowledge. That includes policy, legal, health, and scientific knowledge that is often stated in natural language text, as well as information integration mappings. As a rule language, Coherent notes that Rulelog is easier to learn and author knowledge in, than other approaches are; and, is flexible enough to encode almost any kind of logical knowledge without requiring that one be a logician or a programmer to do so.

During a presentation at the DATAVERSITY® Smart Data 2015 Conference, Coherent co-founder CEO and CTO Benjamin Grosof discussed the power of smart rules and deep reasoning with large graph data to unlock business value from data analysis. “Many enterprises and society in general has a knowledge management problem,” he said. There are plenty of technologies for handling the capture and dissemination of data, “but not so many for making effective use of it – analyzing it to make decisions, answering questions, doing monitoring and alerting and explaining.”

Humans need sophisticated automation of analysis that enables them to gain explanations and check findings, debug and trust query responses, as well as ask drill-down questions at enterprise speed and scale. Ergo supports that. But the capabilities are often unaccounted for by current tools that he says are inflexible and opaque when it comes to the questions that can be answered – not to mention shallow and too disconnected from actual decision making and lacking the means to leverage the expertise of humans who know information that’s not in the data and can be complex to say. Smart data rules with deep reasoning, however, can capture subject matter experts’ insights directly and semantically.

Grosof co-founded Ergo as the latest step in a career that encompasses pioneering semantic technology and industry standards for rules combined with ontologies, their acquisition from natural language, and their applications in various sectors. He also co-founded RuleML, holds leading positions in W3C RIF and OWL-RL, and is a senior research program manager in advanced Artificial Intelligence at Vulcan, among other roles. The evolution of enterprise knowledge management that is based on “smart rules for smart data,” as Ergo’s tagline goes, follows this path:

  • The smart rules work with smart data as input assertions
  • The smart rules infer smart data
  • The smart rules are fully declarative/semantic and themselves become rich data, not procedural code
  • The smart rules overall leverage investments in smart data (such as graph data, linked data and ontologies that extend existing relational databases and other more traditional data stores).

According to Grosof’s presentation, companies Ergo has worked with in the financial, e-commerce, and defense sectors have realized as a result agility, flexibility and ease of authoring, and fast updating. They also have gained high accuracy and transparency; explanations and provenance; lower risk of non-compliance or confusion; and, more cost-effectiveness thanks to bringing subject matter experts more fully into the analytics loop. The unique feature of providing fully-detailed, step-by-step, automatically generated query answers, much like a legal argument or geometry proof does, with interactive navigation presented in real time in English, with provenance links to source documents and context, makes it possible for business users, such as compliance officers, to review and check for accuracy or missing information, he explained.

Turning to Textual Terminology

An organization working with Ergo’s platform could take advantage of its capabilities to import RDF data in bulk into the rule engine’s knowledge base, and externally query triple stores via SPARQL, dynamically during overall rule reasoning. It could also handle frequent updates to asserted rules/facts, while maintaining validity of inferred facts:

“People tend to think of data just as input, but the issue with change is how it ripples through and doing dependency tracking or updating efficiently when input assertions change, especially more powerful expressive input rules or knowledge statements,” he said.

When it comes to facilitating a SME’s understanding of rules and contributions to rule authoring, by supporting close relationship between text (English natural language) and logic (rules, queries, answers, and explanations), Textual Rulelog comes into play. “Pretty much anything you can say in English you can say precisely and directly in Textual Rulelog,” he said. It is an implementation of major research advances in logic – i.e. Rulelog – and how to map between logic and English – i.e. Textual Logic. “A key concept is textual terminology,” he noted, with the core of that being that a natural language English phrase would correspond one-to-one to a logical term in Rulelog if the business wants it to, as well as a logical functor in Rulelog. “It’s a useful methodological option, especially for domain knowledge from SMEs,” he said.

As an example, he discussed a compliance automation pilot project undertaken with the EDM Council Financial Industry Consortium with parties including the Enterprise Data Management Council, SRI International, Wells Fargo, and the Governance, Risk and Compliance Technology Centre. A task was to analyze whether a transaction was subject to Regulation W, which limits the dollar amount of transactions that can occur between banks and their affiliates in order to limit risks to the financial system. Three key questions had to be answered: Whether the transaction’s counterparty is an affiliate of the bank; if the transaction contemplated was a covered transaction; and, if the amount of the transaction was permitted to know if it was subject to Regulation W.

The project leveraged EDMC’s FIBO (the Financial Industry Business Ontology), an emerging common language for automation of business processes across the financial industry, to create semantic business rules capable of deep reasoning. It began with the text of Regulation W as written in the Federal Reserve, and then created a knowledge and rule base to capture the regulatory rules, queries, and underlying facts in a simulated example in order to ask natural language queries; follow questions through to justified answers; extend the explanations with as much detail as the user required; and, successfully gain answers to the three key queries to determine whether the transaction was subject to the regulation. A paper about the effort that Grosof contributed to notes that:

“The PoC very successfully reached its goal of demonstrating how a representative subset of RegW could be effectively automated using Textual Rulelog while also leveraging the EDMC/OMG Financial Industry Business Ontology (FIBO). The business benefits of our rule-based solution were dramatic, including: higher accuracy; solution development that has lower cost, greater agility, higher reusability, and more SME participation; and greater scope of automation including SME-understandable explanations/provenance.”

The benefits of the smart data technology from a business viewpoint, as evidenced by this and other projects, are major. You become “more accurate and agile in a way that has strategic implications – not just operating implications,” Grosof said.

Leave a Reply

We use technologies such as cookies to understand how you use our site and to provide a better user experience. This includes personalizing content, using analytics and improving site operations. We may share your information about your use of our site with third parties in accordance with our Privacy Policy. You can change your cookie settings as described here at any time, but parts of our site may not function correctly without them. By continuing to use our site, you agree that we can save cookies on your device, unless you have disabled cookies.
I Accept