| 2006 Technology
Symposium > Information Management
Information Management
Information Management focuses on technologies and processes that enable
the organization, creation, management, and use of information to satisfy
the needs of diverse applications and users.
Automating Metadata Construction to Enable Enterprise Discovery Service
Victor Perez-Nunez, Principal Investigator
Location(s): Washington and Bedford
Problem
Many believe that providing an accurate portrayal or description of the information content using metadata will improve search efficiency and discovery of information items. However, expecting users to manually create metadata for their information resources is not an effective strategy to improve the performance of a search engine.
Objectives
This project will develop methods to automate the creation of metadata that supplement the content metadata in order to improve recall and precision of the search engine's indexer (e.g., Google).
Activities
We will survey approaches for information extraction, select candidate algorithms to combine, and integrate them as a pre-processor module to output contextual metadata. We will extract metadata from the documents available in employees' transfer folders on the MII, and will compare the metadata to the items' content. We may also use our collection of current news for the same objective.
Impact
Expected impact and transition opportunities focus on the network-centric enterprise service of discovery. Our approach will not hamper the use of COTS products such as Google-like search engines, but will improve their recall performance. In addition, the results have broad applicability to discovery for DISA and for other mega-enterprises such as DHS and the IC.
^TOP
Combining Data and Knowledge in Graph Analysis
Dave DeBarr, Principal Investigator
Location(s): Washington and Bedford
Problem
The need to efficiently extract patterns from large graphs is increasingly recognized by many MITRE sponsors. However, fundamental operations such as exact graph matching are NP-complete and therefore computationally too expensive for even moderate-sized graphs. Thus, more complex analyses, such as approximate graph matching, frequent subgraph discovery, or graph classification, are currently impractical for our sponsors.
Objectives
The goal of this project is to develop and empirically evaluate the use of domain knowledge (DK) to improve inexact graph matching, frequent subgraph discovery, and graph classification. We are also interested in better understanding the characteristics of the graphs our sponsors deal with and in making those characteristics known to the external community.
Activities
To investigate the use of DK in graph analysis we will select datasets, develop metrics for characterizing graphs, and design approaches for representing and storing graphs and DK, as well as develop and evaluate algorithms. Our goal is to have a set of tools for supporting a wide range of graph analysis tasks by the end of this project.
Impact
For our sponsors, performing these operations efficiently for large graphs can mean detecting undiscovered criminal organizations, identifying key criminal organizers, and saving millions of dollars previously lost to fraud.
^TOP
Improving Trustworthiness of Enterprise Data for Decision making
David Becker, Principal Investigator
Location(s): Washington and Bedford
Problem
A primary assertion pertaining to decision making is: "Data of unknown quality is inherently untrustworthy. Conversely, data of known quality can be treated appropriately in the decision-making process." Unfortunately, today there is no systematic approach to representing, measuring, capturing, and using information about data quality. Too often, decisions are based on low-quality data.
Objectives
We will use emerging Web tools (XML, RDF, OWL, etc.) to provide decision makers a more complete view of available data. The research will create ontologies and semantic rules for data quality, and add-ons or plug-ins for commercial tools, that allow data quality to be automatically factored into decision making.
Activities
We will study this problem by modeling the semantics of data quality and decision support. We will survey various data quality and decision support tools and techniques and target further experimentation to areas not adequately covered. We will create a prototype environment that uses our semantic constructs and commercial tools and which demonstrates the advantages of using these techniques in targeted scenarios.
Impact
Our work has the potential to reduce dramatically the number of situations where decision makers must act on data that is incomplete or questionable for various reasons. Our results will clarify the relevance and limitations of a data set to support a given decision and suggest ways to improve how data quality information is incorporated into the decision making process.
^TOP
Information Services Prototypes: Social Bookmarking, Semantic Web, About Me
Donna Cuomo, Principal Investigator
Location(s): Washington and Bedford
Problem
Come see the new information services being explored and piloted by CI&T and our partners, and provide us with your input! You've heard of del.icio.us and flikr on the Internet? Check out MITRE's new pilot social bookmarking capability, called onomi. onomi lets users bookmark interesting Web resources and tag them with various keywords. This both helps users to organize their bookmarks and allows the bookmarks to be shared with others. Imagine being able to access the bookmarks of experts in various areas to help you jump start your research! Also learn about the work being done in applying Semantic Web techniques to enterprise situational awareness. Staying ahead of the pack is a key challenge with the explosion of information sources - applying Semantic Web concepts to automatically find information of interest to you helps keep you ahead of the curve. See the MITRE Enterprise Situational Awareness (MESA) demonstration and learn where we are going in this area. And don't miss a sneak peak of our upcoming "About Me" pilot.
^TOP
Laboratory-Epidemiology Information Integration with CDC
Jordan Feidler, Principal Investigator
Location(s): Washington
Problem
Researchers at the Centers for Disease Control and Prevention (CDC) pursue parallel investigations: epidemiological analyses are performed to discover the determinants of disease risk and biological analyses are performed to characterize the pathogen. To date, the integration of laboratory and epidemiological data within an information infrastructure has proved elusive; this limits the public health and scientific value of the data.
Objectives
We seek to develop an information infrastructure that seamlessly links laboratory and epidemiological data. We will develop a proof-of-concept system that tightly integrates disparate data types related to an individual disease, as well as providing interoperability across multiple diseases. We hope to demonstrate that this system enables new capabilities and dramatically increases the value of existing and prospective CDC data.
Activities
We will work closely with CDC to create extensible schemas and UML data models for disparate data types. We will coordinate between the National Cancer Institute (NCI) and CDCs National Center for Infectious Diseases (NCID) to determine how best to adapt NCI software to CDC challenge areas. We will determine the capabilities of alternative data messaging and integration engines strategies.
Impact
In addition to working across government agencies to bring together related activities going on at NCI, NCID, and the CDCs National Center for Public Health Informatics, we will extend our understanding of information and data interoperability, bioinformatics and scientific visualization, and the Semantic Web in the context of a practical real-world problem that has great national security significance.
^TOP
Netcentric Data Sharing
Len Seligman, Principal Investigator
Location(s): Washington and Bedford
Problem
Netcentric data sharing aims to provide data visibility, understanding, and interoperability beyond traditional stovepipes. However, it is insufficient to simply post all schemas to a metadata repository, given heterogeneous semantics. We also need accurate intersystem mappings -- among implemented systems, from those systems to community ontologies, and among overlapping ontologies. Improved techniques for developing and maintaining these mappings are essential to achieving greater agility.
Objectives
We seek to advance the state of the art in managing intersystem mappings, with an emphasis on netcentric environments. Specifically, we are developing (1) a linguistically-intelligent schema matcher, which identifies candidate correspondences across different systems' schemas, (2) an open framework for integration of schema matchers and other data integration tools, and (3) techniques for managing data sharing agreements between data producers and consumers.
Activities
We developed the Harmony schema match tool and are refining and using it with customer schemas. We developed the Integration Workbench framework for combining multiple integration tools and will demonstrate its use with Harmony and a commercial tool that generates data transformation code. Finally, we created a language for specifying data sharing agreements and are developing the DASH tool for managing them.
Impact
The research directly addresses two of the DoD FFRDC's strategic outcomes: "Integration and Interoperability" and "Enabling Netcentric Operations." In addition, the need for greater agility in data sharing is common to all our sponsors. We are advancing the state of the art, influencing the research and vendor communities, and using our lessons learned to advise several customers on improving data sharing.
^TOP
Ontological Loose Couplers for Disparate Information Systems
Marwan Sabbouh, Principal Investigator
Location(s): Washington and Bedford
Problem
We are attempting to scale our techniques for integrating disparate information systems to an enterprise level. To that end, we aim to show that our approach is scale-free, meaning that the unit of work to integrate a system is independent of the number of systems in the enterprise. We will attempt to do that using various strategies that may include definition of shared semantics and inferencing.
Objectives
Our objectives for FY 06 are to discover two or more context ontologies (or loose couplers) in addition to Time and Position and represent them in OWL\RDF, implement techniques to harmonize multiple communities of interest (COIs) across services by using the context ontologies and shared semantics, develop prototypes of net-centric enterprise mediation services, and apply the above capabilities to harmonize multiple COIs.
Activities
We will embark on the following set of activities: mine data models in order to identify the loose couplers, ensure that our techniques meet the scale-free criteria, prototype an automatic capability that translates a data model into Cursor on Target (CoT) schema using the loose couplers identified, transfer the techniques to AF programs by engaging the various wings, and transfer the techniques to academia and industry.
Impact
This research will impact all wings' data strategy and various AF programs, as well as various COIs. We have partnered with a number of programs and have proposed approaches to harmonize the COIs. The loose couplers approach is supported by MITRE. We are engaging researchers at MIT and collaborating with them on tasks related to business rules.
^TOP
Sensor Data and Analysis Framework
Don Landing, Principal Investigator
Location(s): Washington and Bedford
Problem
Current querying techniques for archive and streaming data are insufficient by themselves to harmonize sensor inputs from large volumes of data. These two distinct architectures (push versus pull) have yet to be combined to meet the demands of a data-centric world. The input of sensor streaming data from multiple sensor types further complicates the problem.
Objectives
The objective is to develop an integrated query capability that simultaneously accesses streaming and archive data sets from multiple sensor types. The research will design and test techniques for incorporating the pedigree of geospatial data and develop an approach that can scale while meeting response times.
Activities
The objective is to develop an integrated query capability that simultaneously accesses streaming and archive data sets from multiple sensor types. The research will design and test techniques for incorporating the pedigree of geospatial data and develop an approach that can scale while meeting response times.
Impact
This research will create an opportunity to transfer knowledge from academia to MITRE, which will allow us to better support our sponsors. The proposed framework will be an enabler for national and tactical management of sensor data. We will help to enhance the Borealis Open Source project to help address our challenges.
^TOP
Standard Rule Language for Enterprise Application Integration
Suzette Stoutenburg, Principal Investigator
Location(s): Washington and Bedford
Problem
To defeat emerging threats, C4ISR systems must be dynamic and adaptable. Separation of rules from executable code supports the ability to dynamically modify system behavior in complex, changing environments. To realize the benefits of rule separation, a Rule Language Standard is required to support the sharing of rule abstractions across domains, thus enabling agility and interoperability.
Objectives
Our goal is to advance the state of the art in the Semantic Web Rule Layer by developing a set of demonstrable recommendations for how the rule language standard should evolve. We will demonstrate how rules can be used for agile management of information flows in complex, dynamic C4ISR environments, allowing identification of DoD requirements for the evolving standard.
Activities
We will explore the interaction between the rule and ontology layers of the Semantic Web and demonstrate how a standard language should best express each in combination. We will examine orchestration of inferencing across layers, rules for dynamic service behavior, dynamic rule distribution, and rule annotation for discovery and reuse. Results will be shared with academia, standards organizations, and sponsors.
Impact
This research will contribute supporting evidence for how the Standard Rule language for the Semantic Web should evolve. The results will advance some of the most critical DoD requirements, including Enterprise Integration, Interoperability and Net Centric system development. The concept of dynamic Service Oriented Architectures will be advanced, supporting our sponsors' evolution to agile, machine-to-machine environments.
^TOP
Velocity
Dave Anderson, Principal Investigator
Location(s): Washington and Bedford
Problem
Intelligence analysis and situation awareness challenges are best addressed by bringing together disparate forms of information, currently with no common access mechanisms. While much information is available on community intranets, finding the relevant information is prohibitively time-consuming. And it often remains "locked up" in analysis tools or behind custom interfaces provided by the data producers, preventing the analyst from fusing it with other data that has been collected or created elsewhere.
Objectives
We will demonstrate the value of aggregating disparate information sources currently available over the network by combining content search with geospatial and temporal search cues. When working a specific problem with specific information requirements, Velocity points the analyst in the right direction. This speeds up analysis by providing relevant results directly in the analysis tool of choice.
Activities
We have constructed a prototype information aggregation service. It currently incorporates three classes of network-ready data sources (text human reporting, structured data, and semi-structured XML) allowing users to find content by geospatial coordinates, time, and full-text search. Results can be browsed directly, or downloaded in generic formats, including an Excel spreadsheet, ESRI Shapefile, or Google Earth KML.
Impact
We are demonstrating the value of a networked information aggregation service to intelligence analysis and situation awareness applications. Analysts can ingest and review the results of their queries directly in their analysis tool of choice, greatly enhancing their productivity. Velocity provides an effective strategy for accessing and exploiting the increasing amount of data available on community intranets without having to ingest and collocate all that data into a single site.
^TOP
|
