A PRA~ATIC CONCEPT OF THEME AND RHEME FOR MACHINE 
T~TION 
Christa Hauenschild 
SFB 99 - E2, Universitat Konstanz, Postfach 5560, 
D-7750 Konst anz 
The concept of theme and rheme we want to propose in 
this paper is pragmatic in two senses of the wordz the theme/ 
theme structure (TRS) is viewed as a strategy of communication 
and thus belongs to pra~natics, and it seems to offer a prag- 
matic solution for a problem of machine translation. 
The starting-point of our considerations is defined by 
several arEuments claiming the ne@essity to preserve the TRS 
of sentences (or utterances) in the process of translation, 
as stated in Koenitz, 1978z 
- the exhaustive-listing reading of the theme 
- the presuppositionsl character of the theme 
- the connexion of the THS with the degree of textual coherence 
. the correlation of themeness and definiteness 
- general considerations in a framework of communication 
theory (readability of texts eto.) 
Further arguments might easily be added to this list 
(e.g. the relation between the TR3 and the scope of operators). 
In principle all of them hold for machine translation as well, 
at least if it is pursued within a theoretical fremework, as 
e.g. in the case of the project SALAT (of. Hauensohild, 1978) 
and of the successor project "Ubersetzungsbezogene Kontext- 
theorie" (Context theory for translation) in the spirit of 
which this paper was written. 
- 121 - 
These considerations lead to the question whether it is 
necessary to compute the TRS of ever7 translated sentence, 
which would be very difficult - if at all possible - Eiven 
the high degree of ambiguity of written texts with respect to 
the TRS. 
We want to propose a somewhat simplified procedure 
which nevertheless ensures the preservation of the TI~. 
Theoretically, this procedure is based on a concept of theme 
and theme that ascribes them a strategic role in communicat- 
ion, controlling, as it were, several other factors (semantic 
and syntactic). If these latter factors are kept constant 
during the tra~slation process, we can be sure that we have 
preserved theme and theme as well (or that the boundary is 
irrelevant in the given sentence), 
The central hypothesis of this paper is the following: 
in most cases it is sufficient to preserve the underlying 
(communicatively conditioned) order of main elements, which 
is conceived in a wa~ ve~J similar to that of Sgall and Ha- 
JiSov~, 1977/78, or Ha~l~ov~, 1980, but without indi@ation of 
an exact boundaz7 between theme and theme. ~he validity of 
this hypothesis is, of course, limited to certain languages 
(we treat German, French and Russian) and to certain types of 
texts, aiminE exclusively at an optimal transmission of inform- 
ation. With such a limita~ion, it is very often possible to 
derive the underlying order in a straightforward way from the 
surface order of the main constituents of a sentence. 
It might seem surprising that this should hold for such 
languages as German, French and Russian, ~hich are normally 
viewed as ezJ~ibiting different types of word-order reEularit- 
lea. But if we look at texts of the mentioned ~ind (e.g. 
reports for international organizations), w@ find that in all 
three languages they axe organized in a similar way, obeying 
more or less to some kind of "linearity principle"s sentences 
- 122 - 
normally start with the elements connecting them to the prec- 
eding context, the elements carrying "new information" are 
situated at the end. Such organization seem- to be most 
effective in the sense of an optimal transfer of information. 
Thus we propose a procedure that derives the underlyin~ 
from the surface order of the main constituents using some 
transformations for special oases (e.g. cleft-sentences). 
The synthesis will yield the corresponding surface order of 
the target language, again with a minimal set of transforn~t- 
ionse 
The procedure is to be controlled by several semantic 
factors that are closely related to the TRS (compare the 
ar&~msnts for the preservation of the TR3 in translation 
mentioned at the beginning), but have to be kept constant on 
independent grounds s 
- anaphoric relations 
- definite or indefinite (or generic) interpretation of noun- 
-phrases 
- scope of quantiflsrs, negation and other operators 

References 

Ha~i~ovd, Evaz A dependency based specification of topic and 
focus Is background and motivation. Journal of Linguist- 
ic Galculus, 19801 1-2, ppe 93-109. 

Hauensohild, Christaz SALAT: System for Automatic Language 
Analysis and Translation. Paper read at the 7th Inter- 
national Conference on Computational Lin@uistice, Ber- 
gen (Norway) 1978. 

Koenitz, Berndz Die Thema-Rhema-Gliederung - lnvariante bei 
der Translation. Linguistische Arbeitsberiohte 19, 
Leipzig 1978, pp. 75-91. 

Sgall P. and HaJi~ov~ E.: Focus on Focus (Part I and ll). The 
Prague Bull. of Math. Linguistics 28, 1977, pp.5-54, 
and 29, 1978, pp.23-41. 
