ON THE TYPES OF SEMANTIC COMPRESSION OF TEXT 
N. No Percova " 
Faculty of Numerical Mathematics and Cybernetics 
Moscow State University, U.S.S.R. 
Any text is constructed in accordance with Implicit 
principles of semantic compression. The paper attempts to 
describe some of these principles. 
There are two main types of text semantic compression: 
(1) the omission of "meaningful" parts, and (2) the omission 
of standard connections among parts of the text. 
1. The first type of compression is subdivided into: 
(a) omission within a predication (by a predication is meant 
the structure- "predicate - P, its aotants - A"), and (b) 
omission of the whole predication. 
&. In special conditions a part of a predication should 
be understood as the whole predication in accordance with 6 
basic rules+/: A -~AP, P -~AP, A...P -~AP, Ai...A j -~ 
-FAIPA j, A i -~AiPA j, A i ~Aj- 
Examples. 
P -~ AP 
The patient was operated. -~ The patient was operated 
by surgeons. 
Ai...A j -~ AIPA j 
After t~e operation Nick feels good. -9 The operation 
was made on Nick. 
+/ In general, as it seems, rules of a text grammar, unlike 
most ruAea or most sentence grammars, should be not obligatory, but optional. 
- 229 - 
b. The omission of a predication is diagnosed by the pre- 
sence of another predication (PRj), which is connected with 
the omitted predication (PR) by such relations (j)'as "ps~t" 
(including stsLrting, intermediate or final point or interval), 
"attribute", "accompazkTing", "reaction", "cause", etc. (the 
number of such relations seems not to exceed 20). In such 
cases the ~nderstanding of the text includes the rule: 
PR~-~. I~R. 
Examples. 
= attribute 
to put the foot in the stoup -~-to mount a horse 
j = reaction 
The Sahalinians take meas~uces to liquidate the consequen- 
ces of the typhoon. The roads and bridges sure being re- 
constructed. -- The roads and bridges were distroged by 
the typhoon. 
2. The omission of standard connections among the parts 
of the text can either be universal or depend on the type of 
the text. The texts can be divided into two main ~p~cups- 
"physically oriented" and "mentally oriented". Roughly speak- 
ing, physically oriented texts describe events as they occ-r 
and ~p~oup the events in accordance with their sequence, while 
mentally oriented texts describe the events as they are re- 
flected in the text author s mind and group the events into 
(individual or socially acknowledged) classes. 
an important subclass of (2) constitute the cases of "de- 
terminating". We call a deteminer a unit which relates not 
only to its immediate context, but to snme larger piece of the 
text, up to the text as the whole. (Determinex~ can be either 
manifested or omitted in the text.) Let s compare two texts- 
(i) Yesterday Nick went to the theatre. Today he is 
going to the cinema. 
(ii) Yesterday Nick went to the theatre. He met his friends. 
On returning home they discussed the pe~ox~n~nce. 
- 2.30 - 
In (i) the meaning "yesterday" relates only to one sentence, 
while in (ll) this meaning relates to the whole text. A piece 
of the text connected with a definite determiner is called 
ItJ "sphere of oper@tion". Along with constant determiners 
(such as one in text (ii)) there are variable determlners: 
their meaning is regularly changed while moving through the 
text (something like a co~ter). An ex~nple of a variable de- 
terminer is a sequence of predicate time markers in physically 
oriented texts, which usually are partially omitted and should 
be reconstructed during the text aualysis. 
To process a constant determiner is to find its sphere 
of operation, to process a variable determiner is, besides, 
to find out a proportional change of its meaning. 
The reconstruction of text omissions demands using dif- 
ferent kinds of information, among which an important role 
belonsm to "c~inon-sense" knowledge. This knowledge is diTided 
into general (typical relations of objects and events) and 
concrete. The concrqte knowledge is necessary, for instance, 
to maderstand such a sentence as Flamin~oes_bonos exploration 
Bhowe that the are ~f the birds is about 20 million years, 
where the linguistic meaning of the word ~ contradicts the 
concrete knowledge of the life limits of creatures, which 
cause the rereading of the sentense. 
Another interesting type of information needed is the 
information about a possibility of a "sub~ectiTe" description 
(from the point of Tiew of either the author or the object 
described) • 
Examples. 
Through the bus window she looked at a rooting building. 
-~ It was the bu but not the building which ,,oTed. 
The people of the Hungarian Tillage of Ekshu are proud 
of an interesting bullding- the chttrch shaping a 
m~hroomo -~ The church is situated in or near the 
Tillage of Ekshu. 
- 231 - 
