HCI & IM R&D develops and enhances technologies to facilitate and improve communication between humans and computing devices. Federal investments in HCI & IM lead to increased computer accessibility and usability and to a greater understanding of computing systems, computer interfaces, and communication networks.

FY 2001 R&D areas include:

• Battlefield robotics
• Collaboratories for knowledge and data sharing, group decision-making, and operation of remote instruments
• Digital libraries
• Information agents for collecting and analyzing data
• Management, visualization, and exploitation of information and knowledge, including large knowledge repositories
• Multilingual document and speech-to-speech translation and understanding
• Multimodal interactions between humans and computer systems, including speech recognition tools, audio interfaces, and haptic devices
• Remote, autonomous agents
• Universal access
• Virtual reality environments

Managing and "seeing"
digital information

NSF's Digital Government program encourages and supports research and interagency collaborative efforts to make Federal information and services more useful and widely available to the public without compromising the privacy and security of either citizens or the Government. Digital Government supports projects that bring computer and IT researchers together with Federal agencies with significant information services missions. The program funds research; workshops and planning grants in areas such as statistical graphics; distributed geographic information system (GIS) image storage and retrieval in field data collection; regulatory compliance reporting; geospatial ontology and multimedia data mining; and coastal management and decision-making.

As part of an ongoing in-depth study of how IT R&D can more effectively support advances in Government uses of IT, Digital Government and NASA funded two workshops and reports in conjunction with the Computer Science and Telecommunications Board (CSTB) of the National Research Council. Findings of a workshop on IT research for crises management are available online at http://www.cstb.org. A second workshop, on February 9-10, 1999, provided an opportunity for IT researchers, IT research managers, and academic statisticians to discuss with Federal managers how they might more effectively collaborate to improve Federal statistics. The CSTB's formal report on these proceedings is also available online.

Other representative Digital Government activities include:

• An autumn 1999 meeting at the University of California-Santa Barbara on protocols for geographic naming in digital gazetteers
• A University of Maryland study, in partnership with the U.S. Bureau of the Census, the Federal Geographic Data Committee (FGDC), USGS, and Brazilian University, to develop new algorithms and systems for indexing, querying, and joining of GIS spatial data with outputs of tabular data (spreadsheets) rather than rendered maps
• A National Institute of Statistical Sciences effort to develop a query system for disclosure-limited statistical analysis of confidential data using metadata, risk reduction and visualization, and query history databases. Research partners include the Bureau of Labor Statistics (BLS), Carnegie Mellon University, the Census Bureau, Kansas State University, LANL, the MCNC computing center, the National Agricultural Statistics Service (NASS), the National Center for Health Statistics (NCHS), the Ohio State University, and the University of Maryland.
• A University of North Carolina-Chapel Hill study-in partnership with BLS, DOE's Energy Information Administration (EIA), NCHS, Syracuse University, Textwise, the University of California-Berkeley, and the University of Maryland-of ways to improve citizen access to Government statistical data with new techniques for metadata integration, tabular visualization and browsing, and natural language processing
• A Stanford University exploration of technologies focused on universal access to IT capabilities. With partners at the Census Bureau, the General Services Administration (GSA), Marconi, and Synapse, researchers work on technologies for the human/computer interface, including speech recognition, eye tracking, and haptic interfaces
• Development of new Digital Government testbeds and education programs such as:
  • A partnership between the Council on Excellence in Government and the Federal Web Consortium to develop a prototype Digital Government resource center and a curriculum for a Digital Government Fellows program
  • A USC/ISI partnership with the Census Bureau, BLS, and NCHS to create a survey authoring and administrative testbed for online information gathering and analysis

NIST's Systems Integration for Manufacturing Applications (SIMA) program focuses on developing interfaces among advanced engineering and manufacturing software systems that enable seamless communication of information throughout extended manufacturing enterprises. Researchers are developing the technological infrastructure to allow engineers around the world to collaborate effectively, manufacturers to reduce time-to-market while producing and delivering products more efficiently, and customers to choose uniquely configured products from manufacturers without paying a premium. Recognizing that scientists and engineers developing new products and processes require access to authoritative information resources for fundamental technical data, SIMA R&D also focuses on applications for collating and delivering critically evaluated data in new and more intuitive ways. The NIST collaboratory activities described in the HCI & IM section are all part of SIMA.

DoD researchers are developing geographically distributed, cost-efficient, versatile, reusable, and adaptable systems to meet education and training goals for DoD's military and civilian workforces--technologies that can be deployed wherever and whenever the need arises. This multidisciplinary effort is aimed at going well beyond current state-of-the-art prototypes that can adapt to diverse learner communities.

The Office of the Secretary of Defense (OSD) University Research Initiative (URI)-supported adaptive learning technology program includes research on human factors leading to the design of effective, efficient, and user-friendly training environments; research on the cognitive and perceptual requirements for effective network-based learning, including the development of cognitive and psychomotor skills; tests of the effectiveness of new distributed training regimes, compared with traditional methods, especially in training in complex decision-making operations, command, control, and communications tasks, and other skills required of modern military personnel; research that develops a scientific basis for understanding how distributed training systems can best build competencies and skills that remain robust and effective under conditions of fatigue or stress; and research on how distributed training systems can assess a wide range of individual differences in cognitive, sensory, perceptual, or psychomotor performance.

NSF's knowledge and cognitive systems (KCS) program supports research fundamental to the development of intelligent machines--machines that can interact in an intelligent manner with humans (computer-aided machine intelligence) or on their own (autonomous intelligent agents). The program's three research components focus on artificial intelligence:

• Knowledge repres entation focuses on characterizing, specifying, storing, and transferring of all types of knowledge, including functional, categorical, structural, and relational knowledge, and forms of knowledge such as auditory, visual, tactile, and kinesthetic.
• Cognitive processing systems R&D investigates the way intelligent systems manage, transform, and use stored knowledge and incoming information to accomplish tasks.
• Machine learning and knowledge acquisition-the study of knowledge discovery and the way it is encoded-recognizes that machines, like humans, need to develop if they are to produce behavior that resembles cognitive functioning.

Some KCS research will explore knowledge representation in machines and studies of cognitive processes, which may be modeled on what is known of human or animal cognition in decision-making, linguistic cognition, machine learning, planning, reasoning, and sensory cognition. Application areas may include design and manufacturing, network management, medical diagnosis, data mining, and intelligent tutoring.

The emergence of commodity high performance hardware and software is transforming the delivery, versatility, and cost of computing. This trend toward numerous, casually accessible computing devices connected by a ubiquitous network infrastructure is enabling more pervasive computing and fostering the development of new "smart spaces."

"Smart spaces"--worker-aware, perceptive, and connected computing environments--emerge when pervasive devices, networking technologies, and information retrieval capabilities are integrated into a functional work environment. A future interactive smart space might include voice interfaces, integrated multimedia, workload computers, wireless networking with personal communicators, language translators, large screen displays, and touch pads to support command, analysis, and design groups.

NIST is developing a smart space environment emphasizing sensor-based, perceptive interface experiments that can identify individual speakers and what they say. Researchers are addressing sensor fusion issues that must be resolved in order to combine visual and acoustic data streams-using signatures from facial images, vocal acoustics, and directional data from microphone arrays-to create a unique "person recognizer." A combined person recognizer, encompassing all individuals within the smart space environment, will provide higher accuracy, allow individual speakers to be associated with spoken utterances in a group, and allow continuous speech recognition to decode words. This technology can provide the basis for a new generation of perceptive interfaces that support collaborative work groups.

NIST has collaborated with both DARPA and NSA to develop and implement a series of benchmark tests to advance computer speech recognition technology, including near-real-time recognition of unconstrained vocabularies in speech typically encountered in radio and TV news broadcasts. For NSA, NIST has developed and implemented test protocols for transcribing conversational telephone-line speech, as well as tests of speaker recognition technologies. New information extraction measures and tests are being implemented in FY 2000. In FY 2001, NIST plans to continue developing metrics and tests for information extraction from speech and for speech-based user interaction.

Today, U.S. military operators are restricted in their access to information. Most often, military operators use voice over radio or formatted text messages to communicate during crises. To help the warfighter improve readiness and response, more flexible, timely, and dependable access to information is needed. In FY 2001, DARPA will continue its work on a new IT capability that allows people to literally converse with computers to create, access, and manage information and solve problems. The challenge of DARPA's Communicator program is to make relevant information accessible to warfighters in an organized, integrated manner that is readily usable by military personnel equipped with untethered information devices. By blending speech recognition technologies--the results of DARPA's past R&D investments-with natural language capabilities, this will enable users to access and create information from networks without a laptop computer, batteries, keyboard, full-size display, and the time spent connecting to a browser. Key research challenges include dialogue management, context tracking, language generation, input language understanding, and hands-free and eyes-free interaction.

NASA's R&D in human factors will develop advanced human-centered information technology, models of human/system performance, principled methods for human factors design of human-centered computing systems, and human factors expertise to address aerospace challenges. In pursuit of this goal, scientists and engineers at NASA's Ames Research Center are conducting research in adapting the technologies of the future to humans in aerospace systems environments.

Virtual environments that offer only visual displays present interactive simulations that can be seen but not felt. The absence of palpable physics, such as the mechanical constraints imposed by real object surfaces or their weight as induced by altered gravity, compromises the accuracy and effectiveness of human manual interaction with these simulations. The mechanical feedback to the sense of touch necessary for realistic manipulation can be introduced into the simulation via haptic interfaces-force-reflecting hardware with appropriate software-to the virtual environment.

NASA and the University of California-Berkeley have built haptic interfaces for arm-scale and finger-scale manipulation. Both machines are based on a NASA-patented kinematic architecture that enables three degrees-of-freedom translations of a grasped endpoint to be coupled to the rotations of three motors fixed to a common base. Because it is composed solely of rigid links and ball bearing joints, this coupling mechanism permits a proportionally large workspace and is stiffer and more responsive than commercially available cable-driven systems. Researchers are implementing dynamic behaviors and textures in these haptic virtual environments for psychophysical studies .

Miniaturized linkage using new kinematics for haptic displays. This machine is a portable desktop unit weighing 5 kg (including power amplifiers) with 4 N sustained force capacity. Its workspace is one-third the scale of the arm device, enough to allow a full range of fingertip motion when engaged in a precision pinch grip with the wrist supported. The machine is also partially balanced, reducing the portion of available actuator or human effort required to support the weight of the mechanism.

Translingual Information
detection, Extraction, and
summarization (TIDES)

The burgeoning international information infrastructure offers unprecedented opportunities for rapid access to and fusion of information from sources around the world. However, the human language barrier remains a critical impediment to effective access to and use of multilingual information, putting the U.S. at a distinct tactical disadvantage when operating in foreign lands and a strategic disadvantage in the international information arena. To help technical and military personnel overcome this barrier, DARPA has funded research on machine translation and algorithms for computer-based Translingual Information Detection, Extraction, and Summarization (TIDES) in FY 2000 and FY 2001.

The goal of the TIDES program is to enable English-speaking U.S. military users to access, correlate, interpret, and share multilingual information relevant to real-time tactical requirements, without requiring the user to have knowledge of a target language. This will require advances in cross-lingual information retrieval, machine translation, document understanding, information extraction, and summarization, as well as integration technologies yielding an end-to-end capability more valuable than these individual components. Achieving this goal will enable rapid correlation of multilingual sources of information to achieve comprehensive understanding of evolving situations for situation analysis, crisis management, and battlespace applications.

Remote Exploration and Experimentation (REE)

The goal of NASA's Remote Exploration and Experimentation (REE) project is to provide a new class of science mission that features fault-tolerant, high-reliability, low-power, and high performance supercomputing in space. These supercomputing systems will increase onboard processing capability by three to four orders of magnitude above what is currently available, permitting more data collection and processing, mitigating downlink limitations, and reducing ground station operations. With onboard supercomputing, a Mars rover, for example, would be capable of autonomous navigation and autonomous geological exploration and experimentation.

Additionally, REE is eliminating the need for radiation-hardening, a task that puts space computers as much as five years behind Earth-bound technology. REE will allow NASA to use COTS computing technologies to ready computing devices for space flight 18 months after the technologies are available.

The robotic battlefield
of the future

The battlefield of the future will demand weapons, unmanned combat vehicles, and communication systems that can navigate, reconfigure, and cooperate autonomously to accomplish time-critical commands. Many of these agents will be built from COTS products, while others, such as bio-inspired microrobots, are currently under development.

DARPA's mobile autonomous robot software project will develop and transition currently unavailable software technologies to program the operation of autonomous, mobile robots in partially known, changing, and unpredictable environments. The program aims to provide new software that will remove humans from the combat, conveyance, reconnaissance, and surveillance processes, thereby reducing danger to troops, lowering manpower costs, removing engineering constraints placed by human physiology, and extending the range of military hardware. Researchers hope to demonstrate autonomous navigation by an unmanned military vehicle as well as interactions between humanoid robots, autonomous vehicles, and humans.

For robots to be fully integrated into the lives of humans in military, commercial, educational, or domestic contexts, they must be able to interact with humans in more meaningful, natural ways. DARPA is funding research to provide new ways for humans and robots to interact.

One potential way to bridge the gap between humans and robots is give them similar bodies. Given the premise that human-like intelligence requires humanoid interactions with the world, DARPA-funded MIT researchers are creating a robot, named Cog, as a set of sensors and actuators that approximate the sensory and motor dynamics of a human body. Cog's computational control is a heterogeneous network of different processors operating at different levels in the control hierarchy, ranging from small microcontrollers for joint-level control to digital signal processor networks for audio and visual preprocessing.

The Cog project's goal is to produce robots that can observe and respond to natural gestures and verbal instructions from a military commander. The commander should be able to demonstrate actions and supply auditory and visual cues to help the robot correctly perceive the instructions. For these imitative learning techniques to succeed, the robot must learn which aspects of the environment it should attend to and which actions it should reproduce.

The DARPA-funded Kismet project at MIT is training a robot head with eyebrows, eyelids, ears, and a mouth to discern social cues such as nodding or eye contact that are crucial in correctly guiding interaction.

Although successes with high-level, human-like modes of interaction are encouraging, some researchers are turning to other animals, such as ants and bees, for new approaches to robot design. DARPA's software for distributed robotics program is funding research to develop techniques for controlling large numbers of simple, computationally limited robots for missions such as deactivating minefields. For example, one DARPA-funded company is developing coin-sized robots that will respond to human direction and swarm targets such as chemical traces or human intruders.

Access Grid

NSF and DOE are participating in the Alliance's Access Grid, an ensemble of networked resources supporting group-to-group human interaction across the national technology grid. For further information, please see page 42.

BioFutures

DARPA created the "Fundamental Research at the [Bio:Info:Micro] Interface" program, aimed at building interdisciplinary teams of researchers from the biology, IT, and microsystems technology fields to address fundamental research issues at the intersections of these three broadly defined areas.

The bio component of the program includes disciplines from the molecular and cellular levels through the organism and population levels. The info component includes the development of theories, algorithms, models and simulations, and scalable parallel and distributed systems. The micro component includes development of sensors, materials, microfluidics, micromechanics, microphotonics, microelectronics, and large-scale systems created from such components. DARPA anticipates that long-term outcomes of this program will include the emergence of new interdisciplinary research communities and new science and technologies that will provide a foundation for revolutionary systems to satisfy future national and defense needs.

DeepView: A collaborative
framework for distributed
microscopy

DOE's DeepView is a scalable system integrating commercial software with unique computational components to enhance scientific problem solving within a collaborative framework. This "channel for distributed microscopy" offers a listing of online microscopes where users can participate in an experiment, acquire expert opinions, collect and process data, and store this information in electronic notebooks. Users can access capabilities including in situ microscopy, recovery of 3-D shapes through holographic microscopy, and image simulation for high-resolution transmission of electron microscopy.

Researchers are using DeepView to study the response of mammary tissue to low-dose ionizing radiation exposure. In the past, researchers have extrapolated the effects of low-dose exposure from the known effects of high exposure rates. Tissue response to radiation, and hence risk, is believed to be a composite of genetic damage, cell loss, and induced gene products. With DeepView, geographically distributed researchers can automate image acquisition and analysis to better understand the effects of radiation exposure and can capture and integrate the data into remote databases to compare data and generate simulations.

DeepView is a scalable system integrating commercial middleware with computational components to enhance scientific problem solving within a collaborative framework. Researchers at several national facilities have used DeepView to conduct dynamic real-time collaborative experiments, such as this in-situ study of a lead inclusion at room temperature.

Distance visualization

A computational framework has been developed by ANL and the University of California to enable online 3-D tomographic image reconstruction using data obtained from remote scientific instruments-such as X-ray sources and electron microscopes-that is coupled to subsequent collaborative analysis. Using a combination of high performance networking and computing resources, parallel reconstruction algorithms, and advanced resource management, communication, and collaboration software, the research team demonstrated quasi-real-time 3-D imaging of samples on an advanced photon source tomographic beamline. Ten minutes after data collection commences, a 3-D image appears on the screens of project scientists at ANL and other institutions. Over the next 20 minutes, the image is progressively refined as more data are obtained. With this technology, scientists have, for the first time, the potential to change experimental parameters in the middle of an experiment.

The system has been demonstrated for data sources in Illinois and Japan with analysis devices located in California and Illinois. End users were in California, Florida, Illinois, and Japan. This research is supported by DARPA, DOE, NASA, and NSF.

Distributed collaboratories
project

A scientific collaboratory environment is composed of many software components that work together to provide information, control, and a sense of presence to the remote researcher. These components include the infrastructures and tools needed to provide remote experiment monitoring and control, multimedia conferencing and telepresence, cross-platform compatibility, interface coherence and usability, and control coordination and conference management. The goal of DOE's distributed collaboratories project is to develop and deploy the technologies needed to enable access to and allow monitoring and control of remote experimental facilities from the home sites of researchers.

Through experience in building collaboratories, DOE researchers uncovered a need for better videoconferencing tools. Existing multicast-based tools are effective for broadcasting content and for one-to-one interaction, but more functionality is needed in the collaboratory environment. A collaborator must be able to activate a videoconference session, invite other participants to join, and carry on useful discussions with everyone participating equally. Additional tools needed include floor control, remote camera control, and remote control of the audio and video tools themselves.

Computer scientists at DOE laboratory LBNL have developed and deployed a system for remote control of video devices used in a videoconference. The system consists of a device server and a client that together let users control the videoconference cameras and video switcher over the internet. The camera control tools are being adopted by the european Laboratory for Particle Physics (CERN) for use in its Internet conferencing software and have been used by the Internet Engineering Task Force (IETF) to broadcast meetingson the Internet.

Real-time collaborative
access to chemical
disaster information

NOAA's Pacific Marine Environmental Laboratory is exploring network-centric, Java-based methods for time-critical problem solving during chemical emergencies and other disasters involving hazardous materials. The initial phase of the project is focused on delivering situation-dependent data over the Internet from 12 source databases and integrating the data through a series of rules. With succesful integration, a process previously run once every two years and in batch mode only will be updated and delivered with the latest data tailored to real-time circumstances as they evolve. NOAA researchers plan to surround this functionality with synchronous collaborative tools, enabling experts from around the Nation to consult while maintaining a consistent shared view of the data. The research team is also working with EPA to ensure applicability and ease technology transfer.

Manufacturing collaboratory

NIST is developing a manufacturing collaboratory to provide manufacturers, distributors, and researchers with structured methods and practices for implementing collaboratory technologies in manufacturing environments. Initial implementation of this collaboratory enhanced research in robotic arc welding, which requires asynchronous and synchronous collaboration support and the use of diverse data formats (such as video, still images, audio, text, database records, and image annotations) across a geographically dispersed community. This research resulted in a collaborative tool prototype with annotation capabilities that synchronize various data streams with indexing relative to time. The next step for the effort is to evaluate the collaboratory technologies in an industrial setting.

Recent advances in shared distributed virtual environments (VEs) and embedded "smart" ubiquitous devices promise to revolutionize the ways we collaborate with people and interface with computing systems. NIST researchers are integrating these technologies and developing and applying tools for evaluating such environments. Applications include an interface to learn about and control devices in NIST's "smart room" and National Advanced Manufacturing Testbed (NAMT).

Manufacturing simulation
and visualization

Simulation systems enable manufacturers to virtually prototype plant layouts, optimize material and component routings, and assess ergonomic factors prior to investing in plant redesigns or new factories. NIST researchers are investigating ways to augment commercial modeling and simulation software systems with programmable human ergonomic models, new human modeling programming, standard libraries for simulated manufacturing resources, integration of simulation systems in distributed computing environments, and validation of simulation algorithms. Researchers are developing natural language interfaces that can help simulate human tasks in a manufacturing operations environment.

NOAA's live access
server (LAS)

Key requirements for distributed collaborations over the Web are the availability of data and common tools to browse, access, fuse, and analyze multiple distributed data sets. To meet this challenge, NOAA developed the live access server (LAS), a robust, extensible, collaborative data server for use by providers of gridded data from a variety of disciplines. Current users include:

• NOAA Satellite Active Archive
• NOAA Pacific Fisheries Environmental Laboratory
• NOAA Pacific Marine Environmental Laboratory
• Office of Global Programs (OGP) Carbon Modeling Consortium
• Centre for Mathematical Modeling and Computer Simulation (India)
• Laboratoire de Météorologie Dynamique du Centre National de la Recherche Scientifique (France)

As the number of users grows, cross-disciplinary data sharing is expected to expand, leading to novel collaborations.

The tele-nanoManipulator

At the University of North Carolina-Chapel Hill (UNC), researchers have developed the tele-nanoManipulator system, allowing scientists to see, touch, and directly modify viruses, nanotubes, and other nanometer-scale objects over advanced networks using remote microscopes and graphics supercomputers in a unified system. The tele-nanoManipulator has been used to measure the rupture strength of DNA, measure the strength of the adenovirus capsid used in gene therapy, and build nanometer-scale circuits with carbon nanotubes. NIH, NSF, DoD's Army Research Office, DoD's Office of Naval Research, and industry sponsors support the UNC tele-nanoManipulator project. The tele-nanoManipulator team is also exploring how researchers, educators, and students anywhere on the Internet can participate in cutting-edge multidisciplinary science. During a visit to a local high school, UNC scientists used the system to enable the students to reach out and "touch" a virus over the Internet.

Researchers at the UNC Gene Therapy Center have used the tele-nanoManipulator to examine adeno virus particles. These particles are used as vectors in gene therapy, and scientists are seeking to understand how they stick to and move around on cell surfaces. A preliminary experiment is shown here, where the particles have been moved on a mica substrate. An area of the mica has been cleared of virus particles; the particles have been pushed together and then pulled apart.

3-D visualization

NOAA inaugurated its collaborative immersive virtual environment testbed during FY 1999. In FY 2000, researchers are investigating new collaborative uses of the ImmersaDesk-a projection platform that "immerses" the user in a virtual reality environment-with the university community. These activities will move from the current desktop Virtual Reality Modeling Language (VRML, software for creating 3-D visual environments) world to larger immersive environments. Such 3-D techniques have already proven useful in a number of areas including data validation, where 3-D representations can highlight outliers and nonphysical attributes of geophysical data sets more quickly and intuitively than traditional 2-D plots.

Visualization and virtual
reality for collaboration
and manufacturing

NIST is partnering with industry to determine how 3-D visualization techniques and advanced collaboration tools can help improve commercial manufacturing processes and factory operations. Researchers are using VRML to interact with and visualize physical dynamics in virtual worlds and create virtual collaborative spaces that enable remote users to participate in design and engineering analyses. In FY 2000, NIST developed VRML "mirror worlds" that allow users to control and monitor the status of real-world devices. An existing camera interface is being extended within a JINI environment to enable additional services that allow users to control and monitor processes easily and intuitively.

Web-based tools for neutron research

NIST researchers are leveraging Internet capabilities to provide greater access to the agency's Center for Neutron Research, a unique national user facility that develops advanced instrumentation for materials research using neutron beams and makes it available to the private sector by developing methods that will allow external researchers to engage in experiments and analysis of data obtained using the facility via the Internet.

Digital Library of Mathematical Functions (DLMF)

In FY 2000, NIST began developing a digital library of certified reference data and associated information for the higher functions of applied mathematics--functions that aid engineers, scientists, statisticians, and others in scientific computation and analysis in areas as diverse as astronomy, atmospheric modeling, and underwater acoustics.

The NIST Digital Library of Mathematical Functions (DLMF) will be published on the Web within a structure of semantic-based representation, metadata, interactive features, and internal/external links. It will support user requirements such as simple lookup, search and retrieval in mathematical databases, formula validation and discovery, automatic rule generation, interactive visualization, custom data on demand, and pointers to software and evaluated numerical methodology. The online resource is expected to increase interaction between mathematicians who develop and analyze special functions and scientists and engineers who use them.

Some of the world's leading mathematicians--from the U.S., England, France, the Netherlands, and Austria--are participating in the project and developing much of the core material. NIST is exercising editorial control as well as developing and maintaining the Web site as a free public resource. The project, funded in part by NSF, is expected to be fully functional in FY 2003. The DLMF is the successor to the classic 1964 National Bureau of Standards (NBS [now NIST]) Handbook of Mathematical Functions and is expected to contain more than twice as much technical information.

Clinical practice guidelines
repository

The Agency for Healthcare Research and Quality (AHRQ), in partnership with the American Association of Health Plans and the American Medical Association, continued its development of a Web-based National Guideline Clearinghouse (NGC) in FY 2000. NGC is a publicly available electronic repository for evidence-based clinical practice guidelines aimed at helping health professionals improve the quality of care provided to patients. More than 500 clinical practice guidelines have been submitted to the NGC by physician specialty groups, medical societies, managed care plans, Federal and state agencies, and others.

Key components of the NGC include structured abstracts about the guideline and its development; a utility for comparing attributes of two or more guidelines in a side-by-side comparison; syntheses of guidelines covering similar topics, highlighting similarities and differences; links to full-text guidelines where available and/or ordering information for print copies; an electronic forum, NGC-L, for exchanging information on clinical practice guidelines; and annotated bibliographies on guideline development methodology, implementation, and use.

Thousands of guidelines will ultimately be indexed, allowing rapid access to key recommendations and assessments on hundreds of topics. Individual physicians and other providers will be able to review and evaluate comprehensive sources of information to assist them with clinical decision-making and patient counseling. Health care systems and integrated delivery systems will use the information to adopt or adapt guidelines in their provider networks, and educational institutions will be able to incorporate NGC information into their curricula and continuing education efforts.

Surface wind analysis system

The surface wind analysis system at NOAA's Atlantic Oceanographic and Meteorological Laboratory (AOML) is a research tool for assimilating and synthesizing disparate real-time weather observations--obtained from across the world's oceans by ships, buoys, satellites, and aircraft, and from coastal locations--into a consistent wind field, providing a significant new capability in weather forecasting and disaster preparedness. NOAA is implementing a prototype database schema. Object-oriented operational research and emergency management products with Web interfaces are being developed, and efficient platform-independent programming technologies are being investigated. Output will be formatted to work with FEMA's hazard loss estimation (HAZUS) wind module in GIS systems used by the emergency management community.

Researchers at NOAA's Pacific Marine Environmental Laboratory are conducting studies to understand how escarpments, ridges, and seamounts affect deep-water tsunami propagation in the Pacific Ocean, and to determine the accuracy and resolution of bottom topography needed for accurate simulation of tsunami propagation. The initial focus of this work is on tsunamis that are generated in the Alaska/Aleutian Subduction Zone and propagate southward toward Hawaii. The December 5, 1997, Kamchatka tsunami is simulated here.

Web-based bioinformatics
databases

Recent growth in the bioinformatics field has generated large sets of biological data on hundreds of thousands of living species, millions of DNA sequences, and tens of thousands of genes. Through its collaborative institute, the Center for Advanced Research in Biotechnology, NIST is developing two Web-based bioinformatics databases-one for molecular recognition and another for macromolecular structures-to help manage these data and advance experimental and computational research in macromolecules.

The molecular recognition database will be useful in drug design and chemical separations. It will contain the structure of pairs of target and reactant molecules, the properties of each, and their interactions and will be accessible via the Internet for submission of experimental results and content queries.

The macromolecular database expands on the Protein Data Bank that NIST has owned since 1999. As a participant in the Research Collaboratory for Structural Bioinformatics, NIST is working to develop, refine, and apply evaluation algorithms to the contents to improve reliability and consistency and to transition to a new database management system.

Akenti: Collaborating in a
security environment

Collaborative technologies aimed at increasing communication between public and private research organizations require adequate and appropriate security that is flexible enough to accommodate the changing roles and requirements of the various participants.

Akenti is a security model and architecture developed at LBNL to provide scalable security services in highly distributed network environments. The goal is to express and enforce an access control policy without requiring a central enforcer and administrative authority. The resources that Akenti controls may be information, processing, or communication capabilities, or a physical system such as a scientific instrument. The Akenti model has been adopted at DOE and NASA.

Controlled sharing: The
secure collaboratory

Researchers at DOE's LBNL and SNL have developed security tools to create a diesel combustion collaboratory that enables controlled sharing of data and resources among various national labs, diesel companies, and universities. With these security features, scientists can safely share their data and laboratories can grant outside access to powerful computing resources and instruments. In the collaboratory, members are identified by public key infrastructure (PKI) certificates, and resources are assigned access policies to control access to the shared resources.

Fingerprint and mug shot
standards

Beginning with R&D in the 1960s in electronically comparing and matching fingerprint and mug shot images across different computerized systems, NIST and the Federal Bureau of Investigation (FBI) have expanded their collaboration to include developing standards and specifications for capturing and storing fingerprint and mug shot images and creating databases for evaluating techniques for processing those images.

As electronic means for exchanging fingerprint and other personal identification data replace the traditional hard copy form of this information, safeguards must be established to ensure the authenticity and integrity of the information. Beginning in FY 2000, NIST will develop a system in which a smartcard containing an individual's fingerprint and digital signature will be used in combination with the person's real-time scanned fingerprint image as a biometric to verify the authenticity of the electronic record's creator. The system will use a registry of legitimate digital signatures and the electronic signing of the transaction to ensure that the record is authentic and the data have not been corrupted since their creation.

Human ID

NIST researchers are conducting research to advance digital video human identification (ID) technologies. As an initial step in developing a baseline face recognition system, NIST developed the first support vector machine (SVM) face recognition system, which reduces recognition errors by a factor of two when compared to more conventional recognition methods. In partnership with DARPA, NIST extended its human ID evaluation effort to include more complex video recognition tasks. In FY 2000, NIST began developing ground truthing procedures for digital video databases for the DARPA human ID program, implementing baseline digital video human ID systems, developing baseline evaluation procedures, collecting a digital video database, and performing computational psychophysics studies in collaboration with the University of Texas at Dallas.

Health care quality and
clinical performance

AHRQ continued its development of COmputerized Needs-oriented QUality measurement Evaluation SysTem (CONQUEST), a tool permitting users to collect and evaluate health care quality measures and clinical performance measures. Clinical performance measures are tools that help assess the delivery of clinical services, such as appropriateness, safety, efficiency, timeliness, and competence. These measures enable comparisons over time within a provider group, among provider groups and organizations, and of performance results with goals. Because the methods must be uniformly specified and applied to all clinicians being compared, detailed specifications for data collection and analysis must first be developed and tested for comparability.

CONQUEST uses two databases, one with almost 1,200 measures from 53 measure sets from a wide variety of organizations and the second containing 57 clinical conditions, linked by a common language.

Evaluating search engines

Since 1992, NIST and DARPA have sponsored the Text REtrieval Conference (TREC), a workshop series that evaluates search engines. A basic TREC task involves searching large amounts of text (around 2 GB) and producing ranked lists of documents that answer a specific set of test questions.

The TREC project has resulted in:

• More accurate text searching for all users
• Demonstrations of the scalability of statistical retrieval methods that provide an alternative to Boolean search methods
• New test collections
• Improved evaluation methods
• A forum for exchanging text retrieval research results
• An incubator for new research in cross-language retrieval (for example, English language questions to retrieve French documents), searching recorded speeches and news broadcasts, and question answering as opposed to document retrieval

The most recent TREC project involved 66 academic, industrial, and government groups from 16 countries.

Software usability testing

Testing business software generates information about qualities such as effectiveness, efficiency, and customer satisfaction. Computer scientists at NIST have launched a project to encourage software usability by standardizing how companies report their software testing and measuring how useful "common format" data are for potential purchasers.

Such standardization could save firms millions of dollars by reducing the lost productivity and huge training costs associated with buying software that is either poorly designed or inappropriate for a specific task. Participants in the project include many of the country's largest software producers and buyers: Boeing Co., Compaq Computer Corp., Eastman Kodak Co., Fidelity Investments, Hewlett-Packard Co., Microsoft Corp., State Farm Insurance Co., and Sun Microsystems Inc.

Web site usability tools

Usable Web sites are critical to industry for increasing e-commerce sales, improving worker productivity, and lowering costs and user frustration. Traditional methods to measure usability, however, are expensive, time consuming, and labor intensive. NIST is developing tools to help Web site designers and developers analyze sites for potential usability problems and to help usability professionals speed up their site evaluations. NIST researchers are working with industry to refine these tools and incorporate them into next generation Web development and usability testing tools.