aspell-devel
[Top][All Lists]
Advanced

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

Re: [aspell-devel] Re: ASpell


From: Gary Setter
Subject: Re: [aspell-devel] Re: ASpell
Date: Sat, 13 Jan 2007 10:41:12 -0600

----- Original Message ----- 
From: "Barry Cavanaugh" <address@hidden>
To: "Gary Setter" <address@hidden>
Sent: Friday, January 12, 2007 9:02 AM
Subject: Re: [aspell-devel] Re: ASpell


> Goodness I looked at this again and instead of making it
simpler I made it
> more complicated!
>
> As I thought about it to handle multiple languages properly you
need a
> switch from single language mode to multiple. "Multiple Simple"
to drop the
> syntactic features and just do basic spell checking and
lookups, "Multiple
> Primary" gives full consideration to the chosen primary
language while
> operating in the simple mode for the second, and "Multiple
Separate" which
> gives the full results for both and assumes it is being called
from separate
> pages or applications. To make your services smarter you will
have to force
> the calling application to be smarter.
>
> If the services are called for multiple languages and do not
send the proper
> requests ASpell goes automatically into its "Simple Mode" or
compatibility
> mode. If the calling app is multi language aware it can then
make full use
> of ASpell's features.
>
> Anyways I will shut up now and stay on the list in "listening
mode". :)
<snip>
---- Reply ----
Hi,

To rephrase, I believe that national language is only one way of
classifying vocabulary. Within a national language, there are
specialized vocabularies for professions, degree of informality,
era and so on. One document may have a mix of vocabularies. It is
useful to have a large base words, including multiple national
languages. However, the larger the base of words, the greater the
chance of missing a misspelled word. Similarly,  if the word base
is huge and all words are classified as equally correct, the
lists of suggestions is going to be large and have a low signal
to noise ratio.
Classifying the word lists of correctly spelled words not just by
national language but by specialty seem valuable to me. As has
been suggested, aspell can then give more then just a binary
correct/incorrect response. It can give a correct/warning/error
response.

In my opinion, when the application connects with the aspell
service it should be able to ask for as broad or a narrow a list
of languages and vocabularies as it needs. But I don't see the
need for "Simple" vs. "Separate". If the application does not
want the complexity of a tristate response, treat the warnings as
errors or as correct does it matter to the service?

Thank you for listening!
Gary





reply via email to

[Prev in Thread] Current Thread [Next in Thread]