Core Knowledge-intensive Computing

For over a decade GITI has been pursuing research projects in the areas of knowledge intensive computing. We have been involved with research projects ranging in intelligent integration of information in late 1990s to current research in machine reading and folksonomy. We have a unique and proven track record of successfully transitioning results of many of the research projects into operational and user environments. GITI is sought by program managers who are interested in developing capabilities for end users. Key elements of our research projects are to aim our work toward satisfying operational requirements and building early alliances with user representatives to ensure early buy-in from users. The following are illustrative examples of the research projects and capabilities that we have developed for DARPA, IARPA, and military Services.

Semantic Web

We have designed and developed key elements of a large research project for semantically based intelligence analysis enterprise workflow architecture. For this effort we developed a uniformly accessible semantic store conforming to an enterprise-wide ontology. We include branching context representation to organize workflow components’ analytical hypotheses. We use logic programming-based, forward-chaining query language for components to access data from the store. We developed a software toolkit embracing all the foregoing to streamline the process of introducing additional legacy software components as semantically interoperable workflow building blocks. 

Ontology Development

We have staff ontologists who have extensive experience with research and development of ontology-based applications such as a semantic discovery and semantic interoperability of enterprise-level government knowledge management systems. Our ontologists have developed reference- and domain-level ontologies for the DoD and IC. They have designed and developed ontologies based on existing taxonomies, controlled vocabularies, and XML schema and utilize a standards-based open architecture approach.

Folksonomies

We are researching and developing novel techniques for exposing the latent semantic structure in folksonomies (online repositories of community-generated metadata tags). We are evaluating strategies for bringing ontological order to the relative chaos of folksonomies.

Knowledge Representation and Reasoning

We have extensive experience with using OWL ontology for data model and a triple store for data storage in W3 compliant RDF format where our triple store supports reification. We have implemented a triple store based on a high-performance Netezza RDBMS and have developed a SPARQL-to-SQL converter. We have used a variety of ontology editors and specialize in using Protégé and a Java based W3 compliant ontology framework called Jena. We use an OWL inference engine ("reasoner") called Pellet.

Machine Learning

GITI has applied machine learning techniques to analyze the interactions of battlefield C4I system operations staff with distributed map data recorded in logs from fielded systems. The goal of this effort was to enable future C4I systems to be anticipatory and adaptive. By examining the data items added by the staff to their common operational picture, we built a model of the war-fighter’s information needs under different circumstances and use this information to anticipate information flow requirements. Using these models, we were able to determine who would likely want to use newly arrived information. This enables systems to automatically deliver new information to the war-fighters’ maps as soon as it becomes available.

Data Mining

We have conducted research on detection of instances of complex structured patterns in large graphically structured databases for data mining in areas as diverse as intelligence analysis, fraud detection, biological structure determination, social network analysis, and viral marketing. We have developed a combinatorial model to study this problem. We conducted experiments with using this model which yielded significant insights into the effect of various parameters.

-back to top-

Agents and SOA Systems

Agent and SOA architectures and integration

GITI has conducted research and development efforts in software agent technology for over a decade. We specialize in researching and developing novel agent-based capabilities. Our focus has been on addressing technical issues with developing a light weight interoperability infrastructure that supports the integration and interaction among heterogeneous software components, including software agents, objects, and legacy applications.  Our research resulted in the development of a powerful framework known as Control of Agent-Based Systems (CoABS) Grid framework that is extremely easy to use and adopt in contrast to heavy weight agent systems such as Cognitive Agent Architecture  (COUGAAR) that requires significant investment to learn and maintain. The CoABS Grid framework is designed to meet the challenges of the military environment, as well as address the heterogeneity among the participating agent research communities. CoABS Grid has been used by over 500 projects and organizations world-wide. Our framework has been successfully integrated with over a dozen different agent technologies from industry and academia. We have successfully transitioned CoABS Grid to a commercial product known as Intelligent Service Layer. The following are examples of our agent and SOA systems research and development:

Agent Development Toolkit

GITI has been researching the development of a common agent development environment and toolkit which would facilitate creating new Battle Command domain agents and provide agent framework interoperability.  The Agent Development Toolkit consists of:

• A toolkit of software to develop, distribute, monitor, and modify software agents.
• A Common Agent Framework based on the Agent System Reference Model
• A graphical user interface that warfighters can use to control the software agents

Cognitive Agent Based Framework

GITI has researched and designed a cognitive agent-based framework (OmniSeer) for user modeling, reuse of prior and tacit knowledge, and collaborative knowledge services. Our research objectives were to:

• “Connect the dots” by helping the analyst create and evaluate links between seemingly disjoint evidence into coherent and connected knowledge fragments.
• Monitor analyst and system interactions and make inferences to create reusable models of the analyst’s interests, preferences, and context.
• Enable the analyst to make tacit knowledge explicit by creating and editing knowledge fragments.
• Automate the process of highlighting novel information related to analysis targets by computing their value and novelty of new information.

Our framework exploited Bayes networks for user modeling and prior/tacit knowledge modeling to enable more effective information retrieval, automated hypothesis generation, and automated hypothesis analysis.

Policy-based Network Management

GITI is a member of multi-national research project that uses software agents to enable the development of a secure information infrastructure. The research objective is to design and experiment with solutions for provisioning of an efficient, robust, secure and interoperable information infrastructure. GITI provides the software agents for policy-based network management and supports end-to-end policy enforcement across independently owned and operated networks, such as those supporting coalition operations or NGOs.

Workflow and Service Orchestration

GITI has developed novel agent-based capabilities such as Composable Heterogeneous Agents for Intelligent Notification (CHAIN) system. Software agents were developed to monitor and analyze disparate information and notify interested parties when critical events occurred.  Agents were used to read daily informational messages regarding High Interest Vessels (HIV) and monitor Navy chat rooms.  Using CHAIN Plans, watchstanders were notified when specific keywords were used in the chat rooms, or when HIVs approached United States ports and/or Navy ships.

These agent-based capabilities enable organizations to deal with increasing volumes of data in a more automated manner, without additional staff.

-back to top-

Test and Evaluation for Research: Scientific Test and Repeatable Experiments

Test and Evaluation Methodology

GITI and its staff have distinguished track record of conducting scientific test and evaluation of emerging technologies for its customers. We provide experienced staff who work closely with program managers from concept definition to post experiment analysis to ensure unbiased technical assessment of complex information systems. Our unparalleled experience in the last decade supporting DARPA with numerous test and evaluation efforts enables us to offer strategies for success both in terms of scientific accomplishments and meeting programmatic objectives.

We have established processes and procedures for test and evaluation of complex information systems. Our methodology and check lists ensure design of experiments that not only meet program objectives but ensure alignment of activities that could yield significant time savings and costs.

We believe critical factors contributing to our success have been to consistently demonstrate effective leadership in navigating through complex conceptual problem areas, to work in a collegial manner with researchers and other integrators, and produce innovative, technology-neutral frameworks that enable researchers to focus on their development and to produce rapid, scalable, and transitionable technologies.

Depending on the program’s needs, GITI typically provides the following services

• a robust system concept technology transition and insertion environment that accommodates competing technologies, enabling the Government to get a balanced view of alternative and competing technologies, including a commercial baseline,
• leverage our role on programs and our involvement with the Intelligence Community and the DoD to identify opportunities for early transition technologies to the user community, to orchestrate meaningful demonstrations of technologies in a real-world setting, and to assist in transitioning those technologies,

• define appropriate component and system level metrics in collaboration with the research community and the functional user community,
• provide instrumentation, monitoring, and visualization services,
• assist in the development and capture of metrics,
• design and conduct Technology Interoperability Experiments and Integrated Feasibility Demonstrations of technologies in an operationally meaningful context,
• deliver a quick-start, demonstration within a few months
• establish a Community Clearinghouse and web portal to enhance information and technology sharing and commonality across the research community

As a test and evaluation performer we often help to focus and accelerate the pace of research projects by developing a clear, program-wide system concept that fosters the research and development of representations and artifacts needed for interaction across program areas.