[unrev-II] Ontology-based Information Extraction & Collaboration Proposals

From: John J. Deneen (JJDeneen@ricochet.net)
Date: Wed Dec 27 2000 - 11:09:44 PST

  • Next message: Henry van Eyken: "Re: [unrev-II] Ontology-based Information Extraction & Collaboration Proposals"

    Interesting proposals to think about, discuss, and use as a benchmark
    for OHS/DKR development..

    265 Dogwood Ridge Road
    Montgomery, PA 17752
    Phone: (570) 547-6774
    PI: Todd Griffith
    Topic#: DARPA 00-013

    Title: Intelligent Adaptive Software Construction

    The objective of this proposal is to show the feasibility of building a
    set of software tools enabling researchers (i.e. scientists, inventors,
    designers, planners, or investors) to represent multiple problem-solving
    strategies. These DiscoveryTools will allow researchers to quickly and
    cheaply automate discovery processes thereby increasing the productivity
    of R&D budgets. The tools will provide a graphical user interface for
    encoding the strategies (i.e. tasks, methods, and knowledge) required to
    solve specific problems, and will allow the users to "run" these
    strategies over sets of data. Most software tools attempt to retrieve or
    display knowledge from databases so that researchers can act on that
    knowledge. This research attempts to show that this can be done the
    other way around. Instead of supplying researchers with knowledge to
    solve problems, supply a program with strategies to act on the knowledge
    to which it has access. DiscoveryTools will allow researchers to specify
    multiple high-level strategies, each of which can be used to solve some
    set of problems. The tools assist researchers in explicating their
    problem-solving tasks as a hierarchy of methods and subtasks. In
    essence, DiscoveryTools will multiply expert researchers leaving more
    time and money available to pursue the most promising options.

    1660 S. Amphlett Blvd., Suite 350
    San Mateo, CA 94402
    Phone: (206) 545-1478
    PI: Ronald Braun
    Topic#: DARPA 00-012

    Title: Ontology-Based Information Extraction from Free-Form Text

    We propose an innovative combination of machine learning techniques
    coupled with a novel end-to-end system architecture built around a
    shared domain ontology to permit ontology-based information extraction
    (IE) from free text. Our Ontology-Based IE (OBIE) system will
    significantly increase end-to-end recall for the IE task while
    maintaining or improving precision. OBIE will accomplish this by
    enabling interaction between different levels of the IE processing
    pipeline simultaneously through a shared ontology. IE components will be
    developed to demonstrate increases in recall permitted by the inclusion
    of hierarchical knowledge in their learning algorithms. Active learning
    and bootstrapping algorithms will be extended to automatically learn the
    ontology of a new domain, to assist in training the IE components, and
    to reduce the burden of annotation on the end-user. Performance metrics
    in a variety of system configurations will allow a characterization of
    performance gains enabled by the proposed architecture. Phase I research
    and development of a proof-of-concept limited prototype will demonstrate
    the feasibility and utility of OBIE's ontology-based IE capability and
    will lay the groundwork for its Phase II implementation.

    600 W. Cummings Park, Suite 3050
    Woburn, MA 01801
    Phone: (781) 935-3966
    PI: Steven Hess
    Topic#: NAVY 00-086

    Title: Metrics for Evaluation of Cognitive Architecture-Based
    Collaboration Tools

    The practice of modern intelligence analysis is increasingly becoming a
    team effort, requiring distributed teams of experts, to collect, filter,
    and collaboratively fuse data into coherent responses to Requests for
    Information (RFI's). To quickly generate a team response, analysts must
    achieve a shared understanding of the problem and the best ways to
    assemble data for effective communication back to a consumer.
    Information Technology (IT) suited to the domain of intelligence
    analysis, will have to support new collaboration strategies that allow
    analysts to represent and share evolving understandings of the world,
    interact with data and link it to emerging arguments, and collaborate to
    combine related argument threads into a single response to consumers.
    The current proposal seeks to blend contemporary theories of team
    cognition with computational modeling techniques and evolving
    collaboration tools to prototype and assess an innovative solution that
    allows analysts to generate collaborative responses through the natural
    process of recognizing and critically evaluating the value of evidence
    to hypotheses about a situation. The resulting tool will help analysts
    focus on relevant data, help them organize data, and provide
    computational tools that insure coherence in final response and assess
    the impacts of evidence on dynamic team understanding. The results of
    the proposed work will be a prototype tool supporting collaborative
    argument construction through the natural process of critically
    evaluating the value of evidence to competing hypotheses about a
    situation. The applicability of our proposed tool can be easily extended
    beyond the targeted domain of intelligence analysis. At the end of Phase
    I we will have a functional demonstration of the tool, and will be well
    positioned to demonstrate the generality of our solution for other
    application domains, both military (e.g., military C2) and civilian
    (e.g., scientific collaboration, journalism).

    4001 Discovery Drive Suite 390
    Boulder, CO 80303
    Phone: (303) 545-9092
    PI: Thomas K. Landauer, Ph.D.
    Topic#: NAVY 00-088

    Title: Dynamic SuperManuals with Latent Semantic Analysis

    The object is to be able to design potentially order-of-magnitude better
    ways to dynamically customize information for given jobs and individual
    maintainers. This project will address three interrelated issues: (1)
    the optimum organization of maintenance-aiding information in text, (2)
    the presentation of the best information in the best order depending on
    the current knowledge and need of the user, and (3) the best handling of
    complex graphics under the constraint of low resolution, small-screen
    technology. In Phase I we propose three activities: (a) Review the
    literature for the last five years on design, usability, usefulness, and
    practical experience with systems for organizing information and for
    aiding and individualizing information finding for maintenance and
    related jobs,
    (b) Create partial prototypes or mock ups to demonstrate the design and
    intent of representative new features, functions, and enhancements based
    on the new computer text-understanding technology, Latent Semantic
    Analysis (LSA), coupled with the empirically proven design, functions
    and features of the SuperBook hypertext manual browser, (c) Propose a
    new overall design for a browser with LSA, enhanced dynamic information
    aids and advanced graphics functions to be prototyped and evaluated in
    Phase II. Maintenance of increasingly complex technological systems is a
    critical and difficult problem for defense, government and private
    sector organizations.
    Traditional print media and current on-line systems are not adequate.
    The expected outcome of this project is identification of the optimum
    organization and presentation of maintenance information in job-aiding,
    and design of potentially order-of-magnitude better ways to dynamically
    organize, present, and customize information for given jobs and
    individual maintainer levels of expertise. A successful technology of
    this kind will be the basis of a high revenue Internet-based service and
    licensing business, and will result in major cost savings and product
    improvements for a wide spectrum of large industries, from
    telecommunications to computer and network systems to airline

    4900 University Square Suite 8
    Huntsville, AL 35816
    Phone: (256) 704-0863
    PI: Andrew Schooley
    Topic#: BMDO 00-010

    Title: Agent-based Knowledge-design Assistance (AKA)

    Over the past decade the growth of data, information and knowledge has
    been accelerating and search engines and simple automation have proven
    to be inadequate at addressing the ensuing information glut. This points
    to the opportunity to apply Intelligent Agent technology to the problem
    by using them as assistants in managing data/information and developing
    the needed knowledge. Our proposed "Agent-based Knowledge-design
    Assistance (AKA) Environment" concept is a significant opportunity for
    the creation of an integrated environment for rapidly formulating
    knowledge bases utilizing agents in conjunction with design pattern
    concepts. The AKA concept provides an environment for hosting knowledge
    design pattern agents, called Template Agents (TA) and using XML as a
    run-time tool for conversion, storage, and maintenance of knowledge. The
    environment presents the user with an integrated view of the available
    TAs using an orchestrating agent, called Design Assistant Agent (DAA),
    which manages, arbitrates and negotiates with the TAs. All the agents
    act autonomously to promulgate their design pattern and agenda within
    the context of the AKA environment and the knowledge base content. We
    believe that the application of design patterns with the AKA environment
    will reduce risk, lead-time, complexity and level-of-effort associated
    with creation of knowledge and management of information. We expect the
    AKA project to be on a FastTrack due to commercialization plans of our
    teammates, Boeing and KnoWave, both of whom have immediate need for the
    technology. The AKA is targeted at the "solution seeking" market which
    is projected to grow from $240M in 2000 to about $1.9B in 2002.

    725 Concord Avenue
    Cambridge, MA 02138
    Phone: (617) 491-3474
    PI: Dan L. Grecu
    Topic#: NIMA 00-002

    Title: Agent for Intelligent Analysis Support

    Intelligence analysts have to respond to the information needs of
    commanders and decision makers leading military forces into potential
    engagements in a diversity of scenarios and in a large variety of
    geographical, political and social contexts. For NIMA this task requires
    the semantic integration of the imagery and geospatial analysis with
    large and diverse data repositories and real-time information. To
    support this information retrieval and integration process we propose
    the development of an Agent for Intelligent Analysis Support. The agent
    will assist analysts by semantically representing the context of the
    analysis problem and using it to support the intelligent information
    retrieval process. The agent will also amplify the information retrieval
    process by using domain ontologies, and will suggest additional
    information sources based on semantic associations with the problem
    context. The proposed approach will enable imagery and geospatial
    analysts to respond to queries anchored in the semantics of concrete
    tactical situations, and to complete their responses with data and
    knowledge available from intelligence channels, technical databases, and
    electronic knowledge repositories. The development effort will comply
    with DII-COE standards and integrate with the toolsets that are
    currently part of the United States Imagery & Geospatial Information
    Systems framework. The proposed effort has significant potential
    applications, both as a technology and as an end-product. As a
    technology the approach will provide tools for context-based semantic
    information retrieval from real-time information streams and
    repositories, in response to the decision-makers increasingly complex
    information needs. Candidate domains range from command and control
    centers, to market analysis, and to assessment in complex medical
    domains. As an end-product, the associated knowledge engineering tools
    will support on-going DoD efforts to develop multi-domain collections of
    knowledge bases, analysis and reasoning tools, that will enable the
    warfighter to take rapid, and informed decisions in complex, real-time

    1911 N. Ft. Myer Drive, Suite 600
    Arlington, VA 22209
    Phone: (703) 841-3500
    PI: Ed Wright
    Topic#: NIMA 00-003

    Title: A Computational Hypothesis Space for Geospatial Information

    This research will develop a revolutionary advance in database concepts
    for geospatial information that will support future intelligence
    analysts in providing responsive support to military operations. The
    concept will provide a computational hypothesis space that provides
    capabilities for advanced reasoning about features distributed in time
    and space based on diverse sources of information. The proposed
    capability will support today's intelligence analyst with the
    challenging task of providing responsive support to military operations.
    When a crisis arises suddenly in an area of the world where little
    initial data is available, the capability will provide rapid access to
    imagery, spatial data, text, and intelligence reports from diverse
    heterogeneous sources with wide variations in currency and quality. Over
    time, as crisis develop, a vast amount of current and detailed data is
    continuously collected and generated by a wide range of sensors and
    production systems. The proposed database architecture will also provide
    capabilities to correlate and fuse all of this information so analysts
    can reason about the military situation and provide a high level of
    support to commander's operational requirements. The revolutionary
    database concepts developed under this research will support advanced
    reasoning with geospatial information in a wide range of military and
    civilian problem domains. Applications include intelligence analysis,
    operational planning, resource management and exploration, and
    transportation and urban planning.

    -------------------------- eGroups Sponsor -------------------------~-~>
    Download Net2Phone's free software to make FREE calls anywhere within
    the US from your PC now! Get great low rates on international calls!

    Community email addresses:
      Post message: unrev-II@onelist.com
      Subscribe: unrev-II-subscribe@onelist.com
      Unsubscribe: unrev-II-unsubscribe@onelist.com
      List owner: unrev-II-owner@onelist.com

    Shortcut URL to this page:

    This archive was generated by hypermail 2b29 : Wed Dec 27 2000 - 11:20:14 PST