[Editor's Note: This week, we welcome Yefim "Jeff" Zhuk of Sallie Mae as he presents a series on Knowledge-Driven Architecture. This series follows up the author’s presentation at the recent international 2011 Semantic Technology Conference San Francisco and further expands on the subject of integrated software and knowledge engineering, originally described by Mr. Zhuk in the book “Integration-ready Architecture and Design.” Part I | Part III | Part IV]

Part II

Looking for a black cat in a dark room

In the corporate world, each clerk and department has their own knowledge compartment.

Prepared for consumption by an author or a single group, information is based on “tribal knowledge” assumptions and naturally has multiple gaps, especially for other groups and departments. In increasingly interconnected businesses, informational gaps lead to productivity loss.

Compartmentalized information is usually hidden and locked inside complex tools. No surprise that we spend from 30 to 50% time looking for information. Not because we love searching… It’s just hard to find something that was hidden (not intentionally!) and especially something that has never been captured.

We often find ourselves looking for a black cat in a dark room.

While looking for a black cat in a dark room, the first thing is to turn on the light. In our case, this means to unlock data, which are hidden behind complex tools, and improve data organization and taxonomy, effectively creating Transparent Enterprise. IT Transparency is one of the most important corporate goals according to Gartner Research [4]. Transparent Enterprise helps to find captured information and reveals informational gaps.

The real fun starts when we can clearly see that the cat is not there…

Changing corporate knowledge culture and finding the cat

To fill in the informational gaps, often in the “tribal knowledge” territory, we’ll have to connect all the dots we placed before and we’ll need both technical and organizational frameworks.

Collaboration of business and IT under an Information Governance umbrella is absolutely crucial in this effort. Information Governance (a natural extension of Data Governance) will go beyond the structured data, to promote common language across the enterprise, guide the process of creating different kinds of content including development documents, business rules and scenarios. Information Governance will facilitate the adaption of policies and methods offered by technical frameworks.

The technical frameworks include semantic information architecture with integrated structured and unstructured data. The key to such integration is a conceptual model (ontology) of a business organization and processes.

Here is the Catch-22. The conceptual model must be based on “good quality” information with no informational gaps. One approach is investing upfront into building ontology. Usually, this involves some consulting help from a knowledge engineering firm. Consultants would interview employees, and turn what they learn into a limited ontology model. This expensive but necessary help is good as a “jump-start”. Generally speaking, managing a company’s information must be one of the core competency areas of a company, not a subject for outsourcing. The biggest problem we face today is how to engage company’s SMEs in the process of sharing knowledge and how to leverage this engagement to fill in informational gaps and complete the model [5]. SMEs are more than busy with their daily routine and will not change their ways… unless you offer immediate benefits.

Conversational Semantic Decision Support (CSDS) helps to solve the problem. CSDS is rather a method than a tool to gradually engage a SME by providing immediate help and increasing efficiency in whatever work is executed daily by a SME. The main tasks performed by CSDS are:

  • Helping a SME with the right questions: providing a conversational script with the direct hints to capture important information.
  • Enabling good answers with semantic support and checking the SME’s entries against Data and Business Models and growing ontology.
  • Establishing a process where the SME’s interactions with the program not only help the SME but also contribute to a knowledge warehouse (ontology).

I’d like to step back and elaborate on several terms and ideas that deserve explanations and examples.

Semantic support, Ontology and Taxonomy …

If you agree with Sir Francis Bacon that “Knowledge is Power”, it might be no surprise for you that there is a science of handling and computerizing knowledge. Ontology or Knowledge Engineering is the science of handling knowledge. The same word “Ontology” also means a collection of related facts and terms.

Semantics is a study of meaning. While processing information, semantic support focuses on understanding the meaning of information and requires a rich informational model.

This is a good point to discuss the difference between Taxonomy and Ontology.

Taxonomy collects the keywords to describe content. Ontology uses more powerful methods to create much richer meta-data models. Besides collecting the key vocabulary, ontology picks up the relationships between the keywords effectively building semantically rich, much more meaningful model of the content.

Let’s say, we have a sentence “Yefim Zhuk was born in Russia”. Taxonomy takes two keywords from the sentence, the name and the place. Ontology would also add the relationships and creates a graph, which represents the content in a more complete fashion. This graph can be extended and later used for querying, in other words for asking questions related to this information.

Ontology helps us elevate information to the next level understood by computers.
The pre-requisites are: High Quality Information (“the cat is there”) and Information Transparency.

High Level or Ontology of Ontologies Multiple ontology modules represent multiple knowledge domains and views.

Integration of modules and resolving possible conflicts can be done via ontology-maps.

Ontology modules can be built gradually in the daily interactions of SMEs with the growing ontology via a conversational interface. Each interaction will work in both directions: helping the SME and contributing to the knowledge modules with SME entries.

Transitioning From “What” to “How” and explaining Conversational Semantic Decision Support (CSDS)

Let’s take a closer look at the provided above example of a conversation between a SME and a machine.

1. A Business Analyst (BA) writes a line of requirements: “application starts with login” and the service would reply “Do you mean the Authentication Service?”

CSDS consumes a line entered by the BA and uses a semantic model to map the line to one of the existing service names. In this case it was the Authentication Service. CSDS will ask for a confirmation, following the rule: a decision has to be made by a person, not a machine, while the machine would take all the boring part of work, like scanning against models and catalogs and mapping to proper terms and formats.

2. The BA would confirm and the service would ask “What Roles and Privileges do you have in mind for your users?”

The machine started a conversation to retrieve important information that otherwise can be omitted. Of course, this conversational script was prepared by someone upfront.

While Business (a SME) knows the answers, an Architect, trained with a systems approach, can provide good questions for the conversational scripts.

An important component is collaboration between a business SME and an architect.
The bridge between Business and Architecture helps us to develop the field and skills of Business Architecture.

3. The SME’s answers are checked against real terms of Data Dictionary, Business Model and growing ontology. This “Reality check” and the following feedback to a SME enable good answers and help fill in the informational gaps in the ontology. If the BA introduces a new Role that is not known to the model, CSDS would ask to define the role via known properties.

This is not an artificial intelligence (yet) but a good combination of a computer’s power with people’s expertise.

The general approach: start small and keep moving, the next step should provide more help to a SME, which in turn will help grow the tree of knowledge and related productivity in a natural way with daily interactions.

Read Part III

You can reach the author of the article at this email address: Yefim [dot] zhuk [at] salliemae.com.

References:

1.      W3C on Large Triple Stores, http://www.w3.org/wiki/LargeTripleStores

2.      Realizing Efficiency & Interoperability: SOA & Semantic Technology in the Business Mission Area (BMA), U.S. DoD, Dennis E. Wisnosky, CTO, DoD (See Mr. Wisnosky’s 2011 Semantic Technology Keynote here)

3.      Integration-Ready Architecture and Design, Jeff (Yefim) Zhuk, Cambridge University Press, A book on Software and Knowledge Engineering

4.      Measuring IT Tactical Spending to Provide Transparency, Gartner Research

5.      Rules Collector, Yefim Zhuk/Boeing, From “tribal knowledge” to rules and rule-based applications


Yefim "Jeff" ZhukDirector of Enterprise Architecture, Yefim leads Information Architecture at Sallie Mae. In the past he consulted government agencies and corporations in SOA and knowledge engineering, shared his expertise at Java One, Wireless One and Boeing Conferences. Cambridge University Press published his book “Integration-Ready Architecture and Design”. In the book and several patents he described a new field of Integrated Software and Knowledge Engineering and Knowledge-Driven Architecture. Hobby: mountaineering and guitar.