Retrieving information from computers is one of the key functions that users perform, but has long been shrouded in convoluted commands. Users have to be able to enter queries in language that the computer can understand, usually from the keyboard, and using Boolean (and, or) logic. With the increase in the number of artificial intelligence systems in business and other applications, this method of retrieving information is proving increasingly frustrating because of the intense training time required to teach new users how to communicate with the computer. Artificial intelligence, whether expert system or neural network, requires that users be able to retrieve information quickly and effectively in order to be useful.
To assist in this retrieval, developers have turned to natural language processing. This type of processing uses everyday speech patterns, words and grammar for commands. Users are able to use such systems more quickly than the command systems currently in use, and the results of natural language processing are promising. There are drawbacks to the systems, however, and they are not yet in widespread use. This research examines natural language systems and their application.
The problem being considered is how to retrieve efficiently the vast amount of information which has been accumulated in computer databases. Currently, the most common method of retrieval involves Boolean logic. In order to conduct a search to find all exempt employees with wages in excess of $50,000, for example, Boolean logic requires that the user enter a query similar to "find all emp exempt = 1 and wage = > 50000" (). Such complex structures require extensive training of users, and are unforgiving. If the user does not enter a space at the right place, or if a comma is used where the computer does not expect one, the query fails. The problem is to improve the interaction and communication between user and computer.