                            SCANNING IN NLP


                            Matthew Probert
                           Servile  Software


Scanning is a process of quick interpretation by ignoring minor words. 
When you scan a written document, you search very quickly for key 
facts. Within the context of natural language processing, the 
principle is the same. 

Minor words include most prepositions. A scan of the sentence, "the 
quick brown fox jumps over the lazy dog." Will reveal "quick brown fox 
jumps over lazy dog." The essence of the sentence remains. It is a 
frequent problem to computers that humans expand their speech with 
superfluous words. An enquirer as a train station may wish to know at 
what time the next train will depart for Manchester. A straight 
forward enquiry would be. "When does the next train leave for 
Manchester?" Is it necessary to specify which train? A scan of the 
question might return. "When train leave for Manchester?" It does not 
sound very natural to a human, but it conveys the same request as the 
initial sentence. 

Writing a computer program to respond to English sentences scanning 
might be used to good effect. The traditional approach, used by SQL 
and Data Ease, is to insist that the operator does most of the 
scanning. These programs may understand phrases such as: "For 
Customers; list records" or "modify records in inventory" Where as a 
human approach may be to command. "Print a list of all the records in 
the customers file." Or "modify every record in the inventory file." 

An imaginary application might be a plain English disk operating 
system which instead of accepting commands like "COPY *.C A:" Would 
understand commands like "COPY ALL .C FILES TO DRIVE A". 

The argument against computer interfaces that understand plain English 
has always been that the operator has to do too much typing. In the 
absence of reliable and cost-effective audio input devices there is 
little defence against this argument. Currently commands to a computer 
must be typed in or selected by pointing to them. 

