[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

duali development



--- Mohammed Elzubeir <elzubeir at arabeyes dot org> wrote:
> because the current duali implementation is not what
> I had in mind for
> duali. Unfortunately the data set required to
> implement the algorithms I have in mind (some of
which are implemented but
> useless without the
> data) is not available and not easy to create.

well, I've been trying lately to understand the way
duali works, because i cant know what u had/have in
your mind to implement, and couldn't really find good
documentation and don't know python, i couldn't really
understand it , but i supposed , after having a look
at the stems, prefixes and suffixes files , i guessed
that it simply tries to find if theses fit together
for some word.

I'm influenced by the idea of a morphological analyzer
that can make a spell checker as a by-product -this
was your note sometime ago- and in addition be very
useful for further analysis,(ie if a grammatical
checker is to be built some time later).

me with two of my colleges, we managed in building a
morphological analyzer that is able to analyze verbs
and some nouns which are -mushtaka- and tell some
possibilities of the input word , being a verb , type
of it such as past, present or imperative, some
information about the suffixes and prefixes, the root
of the words stem and its "wazen", and of course
checked if those prefix,stem,suffix go along
together.-it will joint the OpenSource and
FreeSoftware sometime soon*- of course a lot of
problem are there, some "elaal" and "ibdal" problems
aren't solved yet, and the system accepts word that
are basically right but are eather not used or have
some conflicts regarding "lozoom" and "ta'di" , and
above all it works only on a subset of the Arabic
language.

sorry for writing too much , but this might* be a
proposal for developping duali or starting a
morphological analyzer, otherwise all I'm saying is,
please we need a good documentation for duali , and of
what the vision is about how it should work.


*it wasnt me alone who coded the analyzer, so im not
the only one to decide, but the idea is simple and can
be implemented from scratch.

Maysara Abdulhaq


		
__________________________________
Do you Yahoo!?
New and Improved Yahoo! Mail - Send 10MB messages!
http://promotions.yahoo.com/new_mail