iGouge

I just asked Siri when my next meeting is (3pm) scheduled, so I have a few minutes to blog on some recent meetings centered around our iGouge Machine Learning Projects.  Siri is a great example of natural language processing and how computers (machines) can begin to become part of our daily life (both at work and at home).  iGouge is working with government labs and systems integrators to promote the use of machine learning technology across several areas of research and development. I am reminded of my good friend and deeply respected ideas of Dr. Marv Langston on his blog discussing autonomy and future computers systems.

If you don’t have an Apple iPhone 4s yet, go get one and be one of the 37 million people that bought one over the last three months.  Siri introduces you to natural language processing and for those not old enough to remember LISP and artificial intelligence (AI), she begins to make us think of what machines can do within our government/federal projects.  By the way, I was a Google Android phone user for 3 years and have to admit that the iPhone 4S is a game changer.

So what is the Gouge Software Development Kit (SDK)?

First what is “gouge”.  I founded the company last year based on a few key trends happening in our world. One is “big data” or the situation surrounding all of us. We have too much information flowing into our computers, data centers, brains, eyes, yet still have to solve every day problems and “filter” or turn this information into knowledge.  In addition, most of this data is “unstructured”.  To add to the situation of “big data”, we also have a focus on computer network defense and cybersecurity.  How do we protect our networks? If we can’t protect our networks, then how do we protect our data?  How do we at least migrate to protecting data that is important to coming up with decisions or meeting our mission(s) or service level agreement(s).

Gouge was a term that I learned of at the Naval Academy. It is slang for “knowledge”  or cutting through the mounds of information to find exactly what you needed to know to pass a test, pass a flight lesson, or get past that mean (pain in the ass) upperclassman.  In the spirit of the apple revolution taking our world, I stuck an “i’ in front of “gouge”.  My wife hates the name (RL Phillips was named after her maiden name and did not sound like gouging out an eye!).

So the Gouge software development kit will be a tool/environment for integrators (both government and industry) to develop key applications around machine learning. Specifically, the technology is termed “biologically inspired intelligence”. It is analogous to a brain that is “unbiased” and can both be trained and untrained.

The technology transforms information into a generalized rule set and processing environment. Gouge is capable of multiple higher-order concept formation, i.e. it can learn new concepts by taking multiple layers of context information into account such as visual information structures in pictures or the whole range of lexical, syntactic, semantic and pragmatic context in text. These information structures are transformed into generalized rule sets, which can then be applied to further input facts.

Unlike other approaches, the Gouge SDK enables machines to learn with or without human supervision. The technology automatically generates a lightweight ontology (look that word up!) that detects all relationships among data elements. Learning occurs at the time data is ingested — so it is very fast compared to other approaches.

When incorporated into products as software or embedded as a chip solution (future release), Gouge enables solutions that feature new functions with large improvements in cost and performance.

The features and benefits of the holosemantic data space in combination with the mathematical approaches (math geek combined with free-thinker) are:

• Self-optimized information processing
• Self-controlled content organization
• Multiple higher-order concept formation (multiple high order co-occurrence)
• Autonomic learning via multiple context recognition (unsupervised machine learning)
• Self-generalizing of learned concepts

There have been several projects around statistical models of machine learning and Gouge takes a very different approach.  The technology is based more on self-adapting or self-learning and focuses on being an un-biased “brain” that learns from the unstructured data (bits) that it is fed.  Today, the technology is focused on textual information (in any language). Future builds will support binary data and motion video.

Early projects have focused on bulding case studies for the technology and helping the customer understand that the  human needs a complimentary partner that can handle and process the large amounts of data that is being fed into our lives, enterprises, and overall decision making.  I look forward to graduate students (Georgia Tech, Naval Postgraduate, CMU,etc) coming up with ideas to make our government more efficient and secure.

Some ideas or thoughts to stimulate the application developers in the audience:

• Analysis of twitter feeds. What knowledge can be obtained by monitoring the 140 character feed. Also, should we look at twitter fields as separated tidbits, or correlate and group them by author, geographic zone, etc.
• Analysis of YouTube comments (Who would think to look at those silly comments)?
• Future analysis of YouTube videos
• Analysis of specific government documents to find duplication or contradictions. This could apply to DOD Policy documents, DIACAP/C&A documents, training documents, message traffic, etc.
• Analysis of resumes or fitness reports to find trends or special capabilities in personnel. Finding trends in people, career paths, and to aid the human resources side of the government.
• Analysis of Facebook post. (My recent cruise aboard the USS Carl Vinson in June 2011 showed me potential issues of allowing wide open use of facebook.  Fully understanding the “mood” of the ship’s crew via facebook post data can be valuable.
• Analysis of open internet websites, blogs, pictures, etc
• Understanding the range of machine learning. When is it best to use statistical models vs. association and self learning/un-learning.
• Gouge applications running on computers that compliment the intel analyst on the watch floor.  This technology does not replace the human analyst, only complement and enhance the human planning and decision process.
• Cuts thru human behavior and known cultural biases.  Supports a joint combined command by adding un-bias to the team of sailors, airman, marines, and soldiers. Makes purple more purple!
• Helps support new concepts and ideas centered around planning and decision making in a more collaborative web 3.0 environment (over traditional command and control).

The role of the Gouge SDK is to be a catalyst to spawn a multitude of ideas, applications, and cost-effective development of solutions by today’s government and industry partnership.

Please send your ideas on possible applications and to learn more about helping you be successful to gouge@igouge.com.