Easier C programming: non-determinism facility
Loading...
Date
1995
Authors
Al-Tahat, Khalid Suliman
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
A sketch of nowadays machine translation landscape would show two main tendencies: the
use of symbolic calculation on structures and that of distance on strings. Rule-based
systems use pattern-matching on trees whereas example-base systems use distance on
strings (texts). Our work finds its place in the implementation of the tools necessary to
build a system which would reconcile both approaches: rule- and example-based.
Pattern-matching (identification) and distances are typical operations in NLP systems and
they can be implemented on a data structure called wood. It is a data structure generalising
strings and trees. Both operations face the problem of non-determinism and recomputation
of results. They then become a motivation factor for building our facilities.
This thesis describes two facilities. The first is the non-deterministic facility developed for
computational linguistics objects. Non-determinism is easily handled on parallel machines,
whereas it is not on sequential machines. Implementing non-determinism on readily
available sequential machines could help in proposing better programming facilities for
parallel processing. So, the task is to simulate, in a programming language for sequential
machines, a non-deterministic behavior. The second is the dynamic programming facility
developed to reduce the computation time of some functions by avoiding recomputation of
results. Using these facilities, any function with two arguments of any type can be declared
as a "non-deterministic" or "dynamic". This is done using predefined macros and
specialised modules.
This facilities ease programming in C by offering two qualities: simplicity and efficiency of implementation.
Description
Keywords
Non-determinism facility