New NIST Lab to Speed Commercialization

For instance, the lab will research and develop tests and test methods for information technologies still in their early stages, such as speech recognition, virtual reality modeling language, digital video, cryptography and mathematical modeling.

"The idea for the new lab started almost two years ago," said Kathleen M. Roberts, ITL acting deputy director. "It makes sense now since IT is becoming a commodity, and government and industry requirements are similar except for security. Previously, CSL focused on government needs for IT, that is, how to make government smarter and help it make better use of IT. CAML was completely internal to NIST, covering areas such as payroll, property and procurement. Now, ITL is shifting to help industry make better products.

"We have projects going on that affect the Internet and intranets, for example, conformance tests for VRML. The current working draft is VRML 2.0. The tests we develop help to ensure that the clients or programs developed by specific companies react as they should to the standards," she said.

According to Roberts, the lab has a lot of interaction with industry on standards committees. "All our work is done with industry's consent. In the case of VRML, the need is for tests and specifications at [the] same time," she said.

Another promising area of research, for example, is speech recognition for individuals who can operate computers only by talking. The challenge today is conversational speech and broadcast speech; speech that is read is already down to 15 percent error rates. Research efforts at ITL seek to determine the effect of background noise on speech recognition in general.

Charles N. Brownstein, executive director of the Cross-Industry Working Team of the Corporation for National Research Initiatives, Reston, Va., said the value of ITL is its availability as a place where industry might from time to time find neutral third-party hosting on a sustained basis, as well as a forum for developing standards.

"IT will get more complex. We are clearly in a period of tremendous technological advance," said Brownstein. "At times, it is difficult to come to a conclusion about standardization in any particular technical area.

"Certainly, my colleagues believe the future of IT markets is way ahead of us. As areas are identified, there will be standards efforts. ITL should be a valuable asset in those efforts."

Ralph Z. Roskies, co-chair of the National Research Council's Panel for Information Technology and the scientific director of the Pittsburgh Supercomputing Center at the University of Pittsburgh, agrees with Brownstein.

"IT is a large driver of the economy these days," Brownstein said. "NIST's role at this point is not so much to lead that effort but to be a facilitator. The sort of thing they might do is conformance testing. Does the purported product adhere to the standards and work?"

Another area where NIST might be of value to industry, he said, is security, which is clearly a major issue to the further acceptance of electronic commerce. Still another is serving "as a neutral testing ground for evaluating the claims of these different parties, for instance, in VRML," Roskies said.

When asked why the new ITL was so long in coming, he admitted that everyone was caught unawares with the rapid development in information. In the past, the two labs, CSL and CAML, were directed more to issues in computer science. Now information is the emphasis.

"Another role [ITL] could play is to provide government information services in a convenient way, making more data available from different sources. If you want a model, a perfect thing for NIST to have done would have been Mosaic (the graphical user interface-based Web browser that propelled the popularity of the World Wide Web). NIST's and ITL's role should be to test and evaluate," said Roskies.

Michael E. Lesk, a chief research scientist at Bellcore, Morristown, N.J., found promising his experience with one ITL program, the Text Retrieval Conference or TREC.

"We develop new technology for searching. Our Latent Semantic Indexing is a new way of finding information that searches on associations between terms. We were able to test our work through TREC, which is a completely unique conference.

"In the absence of TREC, you would have various competing claims about the performance of search engines, most of which would probably remain unconfirmed or unresolved," said Lesk.

Detailed information on the thrust and focus of ITL programs is available at http://www.itl.nist.gov/.