|
Information Access |
Division Contact: Martin Herman Smart spaces will make use of numerous aspects of pervasive computing,
which is a shorthand term for trends toward numerous, casually accessible,
often invisible computing devices. These devices are frequently mobile,
embedded in the environment, or connected to an increasingly ubiquitous
network structure. We have developed an integrating platform designed
to promote interoperability, data transport, and distributed processing
that should make it easier to bring these many products together. The NIST Smart Space Laboratory provides a testbed for the research community to bring next-generation technologies together for integration and performance testing in a vendor-neutral environment. The testbed is intended to investigate integration and interoperability among multiple sensors, devices, and computing components, collect test data, and develop metrics for component and end-to-end technologies. Our data flow software makes it easier to integrate components that were not designed to work together, such as speaker identification and speech recognition systems. A group of industrial advisors provides early feedback on our efforts in measurements, interoperability, and standards. The project seeks to bring technology provider companies together with technology consumers, research organizations, and government agencies. Contact: Vince
Stanford Recent advances in automatic speech recognition (ASR) technology have enabled the development of automatic dictation and spoken language understanding systems and prototype spoken language interfaces to information technology. Interdisciplinary efforts involve the natural language processing, information retrieval, and cognitive engineering (for example, usability testing and human factors) communities. Much of the relevant research is corpus-based and relies on shared use of spoken language databases and standard test methods. We work closely with researchers, other government agencies, such as the Defense Advanced Research Projects Agency, and the Linguistics Data Consortium in the collection, processing, characterization, and distribution of spoken language corpora. NIST has produced approximately 980 CD-ROMs as Standard Reference Data for use within the ASR and spoken language research community. NIST also develops and implements periodic benchmark tests to define the state of the art for research ASR systems. These tests include several forms of large vocabulary continuous speech, including speech read from prepared texts, conversational speech, goal-directed spontaneous speech, and, most recently, radio broadcasts. Our areas of interest include characterization of spoken language data, ASR, natural language understanding and information access, and usability testing. Contact: John Garofolo We are working to accelerate the transfer of better text search and retrieval technology into commercial systems. One approach is to conduct a conference attracting international participation from more than 100 research groups in retrieval technology, both from industry and academia. The Text Retrieval Conference (TREC) began in 1992 and is now starting its ninth year. Participating groups work with large, diverse test collections built at NIST, submit their results for a common evaluation, and meet for a three-day workshop to compare techniques and results. Recent research topics for the TREC have included cross-language retrieval, question answering, document retrieval on web data, and spoken document retrieval, among others. The conference serves as a major technology-transfer mechanism in the area of text retrieval. We also have developed a public domain prototype retrieval system-the PRISE system capable of handling over 3 gigabytes of data. This system uses natural language input and state-of-the-art statistical ranking mechanisms. The prototype serves as a research vehicle within NIST and as a starter kit for groups outside NIST interested in working with these new types of search engines or with the Z39.50 (ANSI/NISO) protocol for search and retrieval. The PRISE server is designed to isolate the search engine from the details of the Z39.50 protocol and to minimize the effort needed to interface the server to natural language search engines other than PRISE. Contact: Ellen Voorhees We are supporting image recognition technologies by developing new image recognition methods, developing techniques for the evaluation of existing methods, and providing technology transfer to the commercial imaging industry. Our current technical areas of work include fingerprint classification, human identification technology, and multimedia access standards. Data sets, evaluation software, and fingerprint classification and verification systems are distributed on CD-ROM. We provide technical assistance in research and development for fingerprint automation and identification and have provided over 450 fingerprint standard reference databases, including a unique database of latent (crime scene) fingerprint images used in developing fingerprint systems. We also help improve biometric technologies by providing databases, performance measurement metrics, and software tools to the human identification community; this allows developers to conduct their own tests and evaluations to improve products. We are providing a measurement and evaluation infrastructure to advance technologies dealing with multimedia access. We work closely with the Moving Picture Experts Group (MPEG) international standards community in developing the MPEG-7 standard for content-based access to audio-visual multimedia. We have developed a DVD video collection for use by researchers in content-based retrieval from video. Contact: Charles
L. Wilson We conduct research to measure the usability of software. The Industry Usability and Reporting Project was initiated at NIST at industry's request to help incorporate usability into software products and develop a means for software purchasers to understand the usability of the products they are buying. Prominent software suppliers and representatives from large consumer organizations are cooperating in this project. Working with industry, we are developing a standard reporting form-the Common Industry Format (CIF)-for specifying usability tests on software products. CIF is designed for use by human factors engineers and usability professionals in supplier companies to report the methods and results of usability tests to customer organizations. CIF also will be extended for reporting evaluation of web sites. A pilot trial will be conducted to determine how well the usability reporting format works and to determine the value of using this format in software procurement. The dynamic nature of the web poses problems for usability evaluations. To address how to incorporate usability evaluations into such an environment, we have adapted traditional user-centered design and usability testing methods to produce the NIST WebMetrics prototype software tools. These tools allow visualization of web usage and facilitate evaluation of web sites usability. Contact: Sharon
Laskowski Visualization and Virtual Reality Environments We are exploring the feasibility of information visualization and computer graphics techniques to access, manipulate, and exchange complex information. In the manufacturing domain, we are collaborating with industrial partners to investigate how the application of three-dimensional visualization, specifically Virtual Reality Modeling Language (VRML), can speed up the manufacturing process. This effort includes the use of commercial off-the-shelf software to create visualizations of factory floor assembly lines and support for parts design and assembly. We are analyzing usability and performance capabilities of a number of virtual environment and World Wide Web visualization tools for these and other related applications. We have created a testbed for integrating physical devices with virtual worlds for collaborative engineering and are developing new user interface techniques for collaborative work and measurements for these techniques. Contact: Sandy
Ressler
Date
created:October
22, 2001 |