COLUMBIA UNIVERSITY RECORD  October 29, 1993  Vol. 19 No. 9

              $500,000 AWARD GIVEN FOR WORK IN COMPUTERS

      Kenneth A. Ross, assistant professor of computer science at 
      Columbia, has received a $500,000 Packard Fellowship for research 
      to simplify asking questions of computers.

      Given by the David and Lucile Packard Foundation, it will provide 
      $100,000 in each of the next five years and will allow Ross to 
      pursue research in query languages for databases, which are 
      collections of information stored on computers.

      The award is the second in two years to a Columbia computer 
      scientist. Shree K. Nayar received the Packard last year for his 
      research in computer vision.

      "It's a very prestigious award and we are very proud to have two," 
      said Zvi Galil, Julian Clarence Levi Professor of Computer Science 
      and chairman of the department.

      The Packard Foundation announced Oct. 5 that 20 young science and 
      engineering researchers at American universities would receive the 
      fellowships, which have been offered since 1988.

   FELLOWSHIP PROGRAM

      The fellowship program supports basic scientific research 
      conducted by talented young faculty members, and encourages 
      recipients to continue productive university careers, instead of 
      opting for potentially more lucrative commercial opportunities, 
      said David Packard, chairman of the foundation and co-founder and 
      chairman of Hewlett-Packard.

      Since 1988, the foundation has awarded 120 fellowships to science 
      and engineering faculty members at 40 universities in the United 
      States. Ross was the only computer scientist among this year's 
      fellows.

      Ross, 28, has developed new logical approaches--also known as 
      query languages--that allow computer users to frame complex 
      questions to computers without telling the system how to find the 
      answers.

   LANGUAGE USE

      Computer scientists call the approach "declarative," in that users 
      declare what information is being sought.

      In the procedural languages used in early computers, and in some 
      still in use, an individual had to give the machine explicit 
      instructions at every step to locate and manipulate information.

      Users of declarative query languages would need less training in 
      how to use computers and would spend less time formulating their 
      questions.

      "The bottom line is that this will make more complex queries easy 
      to do," Ross said. He believes declarative query languages will be 
      commercially available in the coming decade.

      Information stored in computers that can be accessed by posing a 
      declarative query is called a declarative database.

      Most databases now in common use are termed relational, in that 
      data is stored as tables in which individual items bear some 
      relation to one another. These systems already incorporate some 
      declarative features, Ross said, but don't allow users to pose all 
      the possible questions they might want to ask.

      The success of a declarative query language lies in whether it 
      allows the user to ask all possible questions and how efficiently 
      it can find the answers, Ross said.

      A good declarative system must be able to translate the query into 
      an efficient execution plan and then carry out the plan, he said.

      For example, a program containing extensive hospital records could 
      discover the number of patients taking antibiotics by scanning 
      every patient's record.

   DEDUCTIVE DATABASES

      Alternatively, the system could create an index of patients taking 
      antibiotics to speed access to the desired records. The latter 
      course of action requires less time and computing power.

      His work has focused particularly on deductive databases, in which 
      information is represented both as raw data and as logical rules.

      One of his principal contributions has been to define a technique 
      called the "well-founded semantics," which has gained widespread 
      acceptance in the field.

      The object of his current research is to develop a declarative 
      approach to object-oriented databases, which are used to construct 
      programs with highly complex structures, such as computer-aided 
      design and manufacturing applications.

   NSF AWARD WINNER

      Ross has completed the first step in that project and has written, 
      with Inderpal Singh Mumick of AT&T Bell Laboratories, a query 
      language he calls NOODLE, or Nonprocedural Object-Oriented 
      Database LanguagE.

      Ross earned the B.S. degree from the University of Melbourne in 
      1986 and the Ph.D. from Stanford in 1991, when he joined the 
      faculty at Columbia.

      He has published 19 research papers in refereed journals and won a 
      National Science Foundation Research Initiation Award in 1992.