[prev in list] [next in list] [prev in thread] [next in thread]
List: aspell-user
Subject: Re: [Aspell-user] Re: Feedback on our approach to Arabic
From: "Ethan Bradford" <ethanb () google ! com>
Date: 2006-03-14 16:40:17
Message-ID: 6b327d700603140840k6d638ba4p3fb45c2a9dcdabc9 () mail ! google ! com
[Download RAW message or body]
It's somewhat beyond my current scope to find out which words are archaic -=
-
I'm just going with Tim Buckwalter's decisions.
However, your suggestion about multi dictionaries can help with another
difficulty for Arabic. In its most formal form, vowels and other
pronounciation hints are added, but normally they aren't. I hate to claim
that the other way is wrong, but can't think of a way to do that with a
general dictionary.
Instead, your suggestion makes me think I can create a formal dictionary an=
d
an informal dictionary, and let the user choose which to use with the multi
file.
On 3/14/06, Gary Setter <setterg@worldnet.att.net> wrote:
>
>
> ----- Original Message -----
> From: "Lars Aronsson" <lars@aronsson.se>
> To: <aspell-user@gnu.org>
> Sent: Sunday, March 12, 2006 7:59 PM
> Subject: Re: [Aspell-user] Re: Feedback on our approach to Arabic
>
>
> > Ethan Bradford wrote:
> >
> > > I don't see having archaic words as a particular problem. It
> > > only reduces quality when a user misspells into one.
> >
> > The words in the dictionary are not only allowed in the text,
> but
> > also used as suggestions. Suppose your English dictionary
> contains
> > both OLD and OLDE (which is an older spelling of OLD). When the
> > user by mistake happens to write OLED, the software could
> suggest
> > "perhaps you mean OLDE?" which will be quite confusing to the
> > user.
> >
> > I scan and OCR a lot of old books in Danish and Swedish, and
> have
> > to build my own dictionaries of old spellings to support OCR.
> I
> > also maintain my own personal aspell dictionary. But I do this
> in
> > two pieces. My main dictionary is for current spellings, and
> then
> > I have a small add-on dictionary that only contains the old
> forms.
> > For OCR I use both, but with Aspell I only use the main
> > dictionary.
> >
> > Every language needs a good dictionary (or two). But then a
> > spell-checker also needs a good way to find the right
> suggestion.
> > This is the real strength of Aspell, at least for English.
> >
> Hi,
> You might look into multi dictionaries. My main dictionary is
> en_US.multi, which is simply a text file with this line:
> add master.rws
> Which adds a read only dictionary. You can have as many add
> commands as you need. You can even add another .multi file.
> You can find more information in the aspell documentation.
>
> Best regards,
> Gary
>
>
>
> _______________________________________________
> Aspell-user mailing list
> Aspell-user@gnu.org
> http://lists.gnu.org/mailman/listinfo/aspell-user
>
[Attachment #3 (text/html)]
It's somewhat beyond my current scope to find out which words are
archaic -- I'm just going with Tim Buckwalter's decisions.<br>
<br>
However, your suggestion about multi dictionaries can help with another difficulty \
for Arabic. In its most formal form, vowels and other pronounciation hints are \
added, but normally they aren't. I hate to claim that the other way is wrong, \
but can't think of a way to do that with a general dictionary. <br><br>Instead, your \
suggestion makes me think I can create a formal dictionary and an informal \
dictionary, and let the user choose which to use with the multi \
file.<br><br><div><span class="gmail_quote">On 3/14/06, <b class="gmail_sendername"> \
Gary Setter</b> <<a \
href="mailto:setterg@worldnet.att.net">setterg@worldnet.att.net</a>> \
wrote:</span><blockquote class="gmail_quote" style="border-left: 1px solid rgb(204, \
204, 204); margin: 0pt 0pt 0pt 0.8ex; padding-left: 1ex;"> <br>----- Original Message \
-----<br>From: "Lars Aronsson" <<a \
href="mailto:lars@aronsson.se">lars@aronsson.se</a>><br>To: <<a \
href="mailto:aspell-user@gnu.org">aspell-user@gnu.org</a>><br>Sent: Sunday, March \
12, 2006 7:59 PM <br>Subject: Re: [Aspell-user] Re: Feedback on our approach to \
Arabic<br><br><br>> Ethan Bradford wrote:<br>><br>> > I don't see having \
archaic words as a particular problem. It<br>> > only reduces \
quality when a user misspells into one. <br>><br>> The words in the dictionary \
are not only allowed in the text,<br>but<br>> also used as suggestions. Suppose \
your English dictionary<br>contains<br>> both OLD and OLDE (which is an older \
spelling of OLD). When the <br>> user by mistake happens to write OLED, the \
software could<br>suggest<br>> "perhaps you mean OLDE?" which will be \
quite confusing to the<br>> user.<br>><br>> I scan and OCR a lot of old \
books in Danish and Swedish, and <br>have<br>> to build my own dictionaries of old \
spellings to support OCR.<br>I<br>> also maintain my own personal aspell \
dictionary. But I do this<br>in<br>> two pieces. My main \
dictionary is for current spellings, and <br>then<br>> I have a small add-on \
dictionary that only contains the old<br>forms.<br>> For OCR I use both, but with \
Aspell I only use the main<br>> dictionary.<br>><br>> Every language needs a \
good dictionary (or two). But then a <br>> spell-checker also needs a \
good way to find the right<br>suggestion.<br>> This is the real strength of \
Aspell, at least for English.<br>><br>Hi,<br>You might look into multi \
dictionaries. My main dictionary is <br>en_US.multi, which is simply a text file with \
this line:<br>add master.rws<br>Which adds a read only dictionary. You can have as \
many add<br>commands as you need. You can even add another .multi file.<br>You can \
find more information in the aspell documentation. <br><br>Best \
regards,<br>Gary<br><br><br><br>_______________________________________________<br>Aspell-user \
mailing list<br><a href="mailto:Aspell-user@gnu.org">Aspell-user@gnu.org</a><br><a \
href="http://lists.gnu.org/mailman/listinfo/aspell-user"> \
http://lists.gnu.org/mailman/listinfo/aspell-user</a><br></blockquote></div><br>
[prev in list] [next in list] [prev in thread] [next in thread]
Configure |
About |
News |
Add a list |
Sponsored by KoreLogic