Magazine article Online

Natural Language Comes of Age

Magazine article Online

Natural Language Comes of Age

Article excerpt

West Publishing Company's announcement of WIN (Westlaw Is Natural) is likely to be remembered as the online event of 1992. It was applauded as the first natural language application in the commercial online environment. West touted the development as the key to the computer novice's ability to use plain English to access Westlaw's massive case law collection. West's full-text legal documents comprise billions and billions of bytes in the legal jargon-riddled database.

There is no doubt that the database is the largest to attempt to use a natural language interface. Having worked with Westlaw since its infancy, I quickly recognized that the introduction of WIN was a giant step in its development. But as a professional, I needed to do more than acknowledge the change. To understand the significance of WIN, I felt like I needed a graduate-level crash course in natural language information retrieval.

I wanted to know, preferably in plain English, how natural language worked and why it worked. To what extent would it obtain superior results over Boolean logic-based searching? What is the likelihood that it will eventually replace Boolean logic as the standard in searching techniques? Who were the developers of natural language? Have other software developers successfully applied natural language technology to large databases? Why did it take so long to develop natural language applications if they are so basic? And what does it all mean to a profession that makes its living in information retrieval?


WIN's method of retrieval, i.e., "natural language," is simple when you get beyond its complexity, but it is not true artificial intelligence. Its retrieval results are often better than those achieved with Boolean logic, but expert searchers don't need to begin sending out their resumes quite yet. Similar software has been developed and implemented on large databases by other companies, and sometimes the natural language applications are more advanced than those of WIN. The timing of the natural language movement appears to have as much to do with hardware as software, and its meaning is . . . well. . .rather profound.

I began reading dissertations and listening to those with natural language expertise. Dr. Turtle, the developer of WIN, and Dr. Koll, the developer and president of Personal Library Systems, both assured me that natural language concepts, which form the basis of their systems, were really quite simple.

Studying natural language immersed me in a discipline whose language is a hybrid comprised of linguistics and mathematics with a dash or two of statistics. It is used comfortably by professionals who have dedicated years of their lives developing algorithms to manipulate a computer to recognize a phrase instead of just a word. It didn't take long until my mind longed for the simplicity of the legalese to which it was accustomed.

I apologize, in advance, to the gifted individuals who work in the field of natural language, if in my description of the forest, I neglect to mention deserving trees. I think Dr.s Turtle and Koll's point was that these events, while revoluntionary in the commercial online setting, are main-stream in the research and development arena.


Natural language was in development long before Westlaw itself was a twinkle in a developer's eye. It is technically a specialty, or subfield in the field of artificial intelligence. It includes members of the, until recently, unrelated artificial language community and the computational linguists. These experts evolved from linguists and computer scientists who had cross-fertilized each other's areas of expertise. To some extent, each group of scientists have continued to develop along their own theoretical bases, but the past five to ten years have seen increased combinations of efforts. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.