58 
SYMPOSIUM PHOTO INTERPRETATION, DELFT 1962 
S.D.C. has been doing research for the past two years in developing a 
system called Synthex which causes a general-purpose computer to accept 
information phrased in everyday English and to respond to questions con 
cerning that information which are also phrased in everyday English. In a 
sense, we have been teaching a computer English [14, 15]. 
Fig. 3 illustrates an overview of this system. It operates in the following 
manner: 1. Information is input to a computer via a Flexowriter. There is 
no restriction as to format or vocabulary. Any everyday English sentence is 
acceptable. 2. Questions are input in a like manner and with equal freedom 
of language. 3. The program searches its entire memory for pertinent answer 
materials and even searches automatically on synonyms of the words used in 
the question. 4. If the memory has no record of a word used in a question, 
then the system will ask the questioner for a synonym and proceed with the 
new word. 5. The answer is output in English and consists of excerpts from 
the stored data. At the present time, the best answer sentence is printed out 
within the context of the paragraph wherein it occurs. If the user so desires, 
the system will also print out information concerning the parent report from 
which this answer was selected, and/or the entire parent report. 
Conclusion 
In conclusion, I would like to report that it is our intention to continue 
and even expand our research efforts in these areas. Generally speaking, and 
with reference to pattern recognition, the following task areas appear to 
warrant early consideration: 
1. The problem of searching photo formats for significant patterns. 
2. Rotation and scale factors in automatic pattern recognition. 
3. Prototype testing on photo materials. 
With respect to photo data manipulation in a computer three other areas 
will be evaluated. 
1. Data collation in a computer and computer-generated narrative summaries. 
2. Keys in computer functions. 
3. The role of free text versus fixed format in photo data input. 
Table 4. A summary of achievable goals 
1. Rapid automatic screening of large volumes of photography. 
2. Recognition of previously identified sites and data storage limited to new facts. 
3. Machine-aided identification of suspect sites. 
4. Computer storage of all information on known sites and man-machine communication in 
unrestricted English. 
5. Rapid access to specific items of information from a single photograph as well as informa 
tion derived from numerous photographs. 
Table 4 is a summary of the capabilities which we believe are achievable 
in a semi-automatic photo interpretation system. I appreciate that the results