NMSU branding

New Mexico State University

New Mexico State University

News Center

NMSU computer scientists strive to ease computer use

Anyone who has ever stared at a computer screen wondering which button to push may be thrilled to know that four New Mexico State University computer scientists are developing software that could make computers easier to use. Even better is the hope that some of these programs will be available in the not-too-distant future.

With help from a $38,000 National Science Foundation research instrumentation grant, NMSU computer science professors Gopal Gupta (front), Enrico Pontelli (back), Janyce Wiebe (standing) and Desh Ranjan (not in picture) have purchased equipment to help them in their quest to create user-friendly computer applications. The equipment includes seven multi-processor PCs, all connected by a parallel network. They are like mini-parallel computers hooked together to make a larger parallel system.

Parallel computing uses many computers together to solve problems that would take too long or are too complex for a single machine. This parallel system is unique, the team said, in that the researchers will use its capabilities for symbolic, rather than numerical, computation.

"You can broadly classify computer problems into two categories," explained Gupta, "number-crunching, or those that require numerical computations, and symbolic, or those that involve reasoning and symbols, like words."

Pontelli said more has been invested in parallel computing for numerical computation than for symbolic problems, although it offers similar advantages in both areas.

Many symbolic problems are intricate and hard to solve, so they take considerable time and resources, Gupta said. Parallel computing can be used to work on larger problems more accurately, he said. The research team will be using the new parallel system for three types of projects -- verifying software systems, developing Internet programming frameworks and cultivating natural language processing.

Using the new equipment to verify software systems is important because many critical applications are run by software, Ranjan said. For example, a nuclear power station using computers to monitor temperatures might initiate counteractive measures or activate warnings if the temperature exceeded certain levels. A malfunction in such a system could be disastrous. Ranjan said the new parallel system will be used to run simulations to help ensure vital software systems run properly.

Developing Internet programming frameworks means creating "smart applications" to allow amateur users to easily program Internet software, Pontelli said. For example, a user might be able to write a program that would automatically search the World Wide Web and then compile the results in a standard format. "The task of developing software for the Internet is currently beyond the capabilities of many programmers, due to the large number of low level details involved in the process," Pontelli said. "The proposed framework will provide a uniform and logical view of the World Wide Web and hide most of these issues." This project has also received funding by the Japanese Institute for Advanced Technology.

Natural language processing (NLP) is the third research area for the new system. "In NLP, we work on enabling computers to understand and speak human languages, such as English and Spanish," said Wiebe, who heads an NLP research group at NMSU. One of the challenges in the field is that human languages are vastly ambiguous, which means computers must be "taught" how to choose the correct interpretation from many possibilities, she said.

A recent $300,000 grant continuation has allowed Wiebe and her colleagues to address such NLP applications as teaching computers to determine which dictionary definition a speaker means when using a particular word. For example, the computer must decide whether "pen" refers to an enclosure for pigs or a writing instrument, Wiebe said.

Another NLP application allows a computer to distinguish facts from opinions. By looking for certain opinionated phrases, such as "unlikely" or "unnecessary," a computer would be able to sort out factual sentences from large amounts of information, Wiebe said.

All of the symbolic computing research is designed to make computers easier for the average user, said the researchers. Whether that means giving the computer instructions in native languages, applying computing to everyday life, or automating tasks like scanning sources to cull out specific information or compiling reports from the Web, the NMSU team contends that users will see these types of applications on their computers in three to four years, if not even sooner. By that time, they also believe public users may have multi-processor PCs on their desktops, given the current trend in technology prices.