From: Sunil M. <su...@at...> - 2007-05-11 05:58:54
|
సోమ, 2007-05-07 at 17:52 +0530 నాడు, Veeven (వీవెన్) వ్రాశారు: > I came to know that there is an aspell dictionary for Telugu > (http://www.swecha.org/wiki/Wordlist_and_Dictionaries). I am trying to > make a Telugu spell-cheking dictionary for Firefox. Since Firefox uses > myspell, I wish to build myspell dictionary. (I started some effort > here http://groups.google.com/group/teluguspellcheck/) > > Is there any way that aspell wordlist can be converted to myspell word list? > Converting Aspell dictionaries to myspell should be very straight forward. The formats are almost one to one mapped according to my outdated knowledge on this. > OpenOffice.org now uses hunspell (which is the successor of myspell > and has lots of other features). And, Firefox would also switch to > hunspell (https://bugzilla.mozilla.org/show_bug.cgi?id=319778). > Ideally, we should come up with hunspell dictionary for Telugu. > Yes, we should switch to hunspell. The current aspell spellchecker is very a simple wordlist without affix rules and suggestion hints. Amba Kulkarni, at IIIT along with a team in HCU has worked on Telugu morphological analyser (GPL, somewhere on http://ltrc.iiit.net/). AFAIK, this consists of most comprehensive set of rules for prefix, suffix and word joining. She told us earlier that a wordlist based spell checker can never achieve high accuracy. We were trying to use this morphological analyser as a spell checker. The code is large set of scripts mostly in perl. We imagined that hunspell with its claims to do morphological analysis will suite our purpose when we feed it the wordlist and rules from Telugu morphological analyser. We have studied hunspell capabilities (which are indeed remarkable) and planned to discuss with her to find out if hunspell will fit the job (perhaps with some improvements). All this is going a bit slow as our team member who is working on this is also working on other things. -- Sunil |