ON AN APPROACH FOR DESIGNING LINGUISTIC PROCESSORS 
Radoslav Pavlov, Galia Angelova 
Laboratory of Mathematical Linguistics, Institute of 
Mathematics with Computer Center, Bulgarian Academy 
of Sciences, P.OvB. 373, I090 Sofia, Bulgaria 
The present paper discusses the principles of designing 
a system for man-computer di~lo6ue in natural, language. The 
system is being elaborated at the Laboratory of Mathematical 
Linguistics at the Institute of Mathematics with Computer 
Center of the Bulgarian Academy of Sciences. 
The desorlbed system requires: 
I. A form~ description of the syntax of basic nuclear 
structures of the natural lanEuage sentences which has to be 
used during the process of syntactical analysis of the users" 
phrases in a natural language~ 
2. Techniques for representing in the computer the 
knowledge about a given problem area. The obtained descript- 
ion has to be used during the process of semantical analysis 
of the users" phrases in a natural language! 
3. A dictionary containing the terms used in the chosan 
problem area. The dictionary is compiled by specialists in 
the given problem area; 
4e A dictionary containing the eel-vice vocabulary (in- 
dependent of a particular problem area) - word-groups as 
"less than", "geater than", "equal to", "slid", "or", "as... 
ash etc. 
5. A basic software system - a data base management 
- 222 - 
system (for example, a relational DBMS), or a program package 
(for example, BMDP - bio-medical data processing). These 
ready software products service users of- a given problem area 
and they usually have their own language for describing and 
processing the data. 
The considered system is viewed as a superstructure 
over the given software system and it performs: 
1. Translation from a language close to the natur~al one 
into the internal software system's language for description 
or processing the data; 
2. An effort to analyse the correctness of the user's 
request in terms of the given problem area description and 
messages when discovering mistakes! 
3. When ambiguities happen a dialogue is carried out in 
order to make the user's request unambiguous. 
4. ~rocessing the results from the basic software sys- 
tem and final shaping of the messages in the context of the 
user's request. 
Such a system is called a linguistic processor (or 
linguistic preprocessor) and is considered as a dynamic sys- 
tem which can be generated over various problem areas and 
various basic software systems. 
By "a natural fan.age as a tool for the man - computer 
dialogue" in this paper we mean the following: 
- a fixed list of appropriate natural language phrases 
is not aseigned~ 
- restrictions are not imposed on the ~rammatical 
structures of the sentences the man enters into the computer! 
- man has to interact with the computer in a correct 
natural language. 
- 223 - 
Dictionary of the terms Structural description 
of the problem area of the problem area 
user analysis of terms I 
Internal \[ 
• level I l 
Dipole In ISem=tical l_l le o, thei case of ~- ~ analysis in ~-~ structural M-- -- 
incorrect --~terms of the ~'~desoription r internal I structural I I J 
structure |description of I ' " \[the problem \[ 
| area \] 
Xnt ernal 1 
level 2 ISoftwsre' 
Generation \[ I Description of \[product \[of an inetruc~-~the particulax Ito be 
\[tion in the ~-Jinternal leuper- particular \[ \[language \[etruct- 
I internal \[ ' lured I 
lan6uage J 
~ Instructions I 
Results 
Output ,to the user L 
Fig.1. Functional relationships between different parts of the 
linguistic processor and the problem area descriptions. 
- 224 - 
The describe approach for creatin~ linguistic processors 
can be qualified as a semantic - syntactical one since semant- 
ical analysis in terms of the given problem area description 
has a leading role in the process of "understanding" of the 
input phrase in a natural language, while syntactical analysis 
has an auxiliary role. The structural description of the prob~ 
lem area is a network designed for the representation of the 
problem area semantics and is viewed as a static, inmoblle 
and determinative tool for "understanding" and accomplishing 
the dialogue. The syntactical analysis does not presume a full 
scanning of the input sentence. It involves analysis of synt- 
actical dependences at different levels with respect to the 
problem area complexity and, naturally, to the input phrase 
complexity. By means of an elaboration of the formal descript- 
ion of Bulgarian syntax, an improvement of the techniques for 
syntactical analysis is envisaged. 
When workir~ out the principles of construction of 
linguistic processors, some peculiarities of the Bulgarian 
lan~age have been taken into account, namely, its flexiona~ 
character, the absence of a case system, free word-order, etc. 
The approach proposed for the realization of the man - 
computer dialogue in a natural language allows to process 
problem areas with different complexity. Linguistics process- 
ors at two levels are bein~ elaborated at the Laboratory of 
Mathematical Linguistics - for access of users - nonprogr~mn- 
ere to a relational data base in Bulgarian l~ags and for 
access of users - nonprogrammere to the facilities for stat- 
Istloal analysis proposed by the program package \]~DI ) 
Bulgarian language o 
- 225 - 
