ON MEANING IN THEORETICAL AND COMPUTATIONAL S~ANTICS 
Jordan Psnchev, Vladlmlr Perlcllev 
Institute for Bulgarian Language 
Chapaev 52, Sofia, Bulgaria 
Theoretical semantics is concerned with the modelling of 
meaning proper, and computational semantics (natural language 
understanding systems, machine translation, dialogue 
systems), with the modelling of communicative meaning, or 
meaning as used purposefully by people in communication ("By 
uttering X with meaning proper Y the speaker intends a commun- 
icative maning Z'). The distinction between meaning proper 
and communicative meaning Is difficult to make since in every- 
d~7 apeech~ which is the only observable manifestation of 
meaning, people are apt to confuse the thing Itself with what 
it is intended to stand for as well as for some other reasons 
(e.g. that both meanings are slmultaneosly acquired In our 
childhood). It is clearly cutj however, in forelgn language 
acquisition. In foreigh language acquisition people learn the 
meanings of words and expressions rather than how these mean- 
ings are used for the purposes of communlcatlon~ since in the 
general case, the letter Is part of their own native language 
knowledge. This explains wh~ we speak that for a Bulgarian to 
have learned the meaning of the English I wa~t to sleep is to 
have learned that in Bulgarian It has meaning equivalents llke 
~skam da sDda~ ~ etc., and not to have learned that in 
communicative acts the former English sentence may be puPpose- 
fully used b~ the speaker wlth the Intention of stating that 
he (she) is willing to sleep, or urging someone to leave hlm 
(her) alone so that he (she) can sleep, and so on. Furthermore~ 
- 226 - 
we can imagine situations In which one knows the communicat- 
ive meanlng of a language expression without knowing Its 
meaning proper. 
Communicative meaning lles at the basis of our intuitive 
under~tandlng of language. Understanding of language crucial- 
1.y depends upon different mental processing accomplished by 
native speakers such as explicating Implicit connections in 
sentences, logical deducing (Schank, Rieger) etc. Strictly 
speaking, this processing does not Involve meaning proper and 
as such As not a part of linguistics. In this sense, the 
claim made In ore.rant semantic theories that linguistic se- 
mantics should e3plaln the In~Itlve understanding of the 
language by native speakers Is not true, unless, of course, 
the notion of llngulatlc meaning Is extended to meanlnglees- 
ness. 
In the paper, some cases of unacceptable conceptions of 
linguistic meaning In the works of semantic theorlsts are 
discussed. 
For further explication of the difference between mean- 
lng proper and communicative meaning (In the sense above) 
classes of amblgultIes are discussed which have different 
meaning proper but still their different semantic representat- 
Ions preserve the same communicative meanlnE. These classes of 
ambiguities share the same (l) referent, (Ii) Implication, 
(Ill) presupposition, or (iv) present communicative equivalents 
In a less well-deflned way. 
In the paper, It Is noted that, In addition to deepening 
Its concept of semantics along the lines of studylng commun- 
Icative mean£ng, as an applied science, computer llnguistlos 
searches for strategies avoiding ~ather than solving some of 
Its most difficult semantical problems. 
:Two such strategies ar~dlscussed. One Is concerned with 
the tackling of the ambiguity between marked and unmarked 
lexlcal items. In the other, It Is proposed that Instead of 
- 227- 
trying to resolve some cases of A. Syntactical ambiguity, 
and B. Semantical ambiguity in the analysis (e.g. for the 
purposes of MT), sentences are synthesized An the target 
languase which are syntactically ambiguous (in the same 
sense), so that the user himself, rather than the linguistic 
anelyser, resolves the ambiguities in question. 
A. John hit the dog with the lor~ bat (Schank) (it ls 
not clear whether wlth the long b~t' is s modifier of hat or 
of do~) ma~ be translated Into Bulgarian as _D3on udarl k~etq 
which preserves the ambiguity of the original 
sentence (udarA -~ ~ or ~ -~ P--~I~MI_;MM~) • 
B. NOW for breakfast we shall want a f ry~n~-u.n (~arrls 
said It was Andl~estlble; but we nearl,y urged him not to bo 
IP._aa~...) (J.K.Jerome) (it is not clear whether for break- 
fast they want to eat the frying-pan Itself or just need the 
frying-pan to prepare their breakfast In it) may be translat- 
ed Into Bulgarian as the s tructura~S_~L~L~ ~ 
edln titan da saC-aim... ( v~p~.~_J~n ~- ~ or 
-~ titan da~zakuslm). 
As an aid to the above-mentioned strategy, a llst Is 
made (within dependency grammar framework) of 36 models of 
structural ambiguity in English and Bulgarian. 
- 228 - 
