WORKING GROUP 1 
BARRETT 
57 
Man-machine communication 
The second area of research, upon which I would like to touch briefly, 
concerns the problem of how to aid in photo information extraction, storage, 
manipulation and retrieval operations. Given something like the data load 
suggested earlier in this paper, it follows that as we reduce the mountain of 
incoming photographs we also build up a new mountain of words which 
contain the information derived from the photographs. We will soon have a 
truly massive library of photo information, and thereby, a new problem arises. 
How do we very quickly get specific answers to specific questions from such 
a library? How can we control the size of this library - let us call it a data 
base — by not storing in it things which are already there? How can we get 
real time (■i.e., immediate), specific help, from the data base in analyzing new 
photographs? Historically, the answer has been to maintain a staff of what, 
in effect, are research librarians, and to accept the fact that immediate response 
is, in most cases, unobtainable. Computers can ease the problem tremendously 
in that they can contain vast amounts of data in digital form, in a small 
physical space, and can search such data at fantastic speeds (a million words 
can be searched in three minutes). Thus far, however, the capabilities of 
computers have not been fully realized in that the user, i.e., the photo inter 
preter in this case, cannot talk to the computer. It does not speak English, 
or French, or German, or any other human language. Questions to the com 
puterized information store and answers thereto are coded, i.e., expressed 
numerically, for digital processing. This requires a pre-edit and a post-edit of 
all correspondence by a human translator, a buffer, with resultant loss of 
time, preciseness, and most of all, the capability of the photo interpreter to 
“browse” in the file in search of his answers.