Antoine Dubourg a.k.a. Tcheko released its latest production : spellchecker.library 1.0
This new library offers spellchecking feature based on a dictionary of words. It uses common and efficient spellchecking algorithms : double metaphone and Levenshtein distance (used in major Office package like OpenOffice, AbiWord...). This first release is aimed mainly at developer as no application currently uses spellchecker.library.
Currently, this library offers '(s)low' performances : spellchecking is done with a simple lookup table. Further version will include faster spellchecking with the help of btree.library to the price of 'huge' memory consumption. Tested with a french dictionary of more than 330.000 words, spellchecker.library took more than 80MB of memory excluding its usage on low end hardware like Efika. Some benchmarking were done to evaluate the spellchecking speed : up to 1 000 words per second was achieved on a Mac Mini@1.42 with the btree version of the library.
Still, further version of the library will be designed to handle low memory case to accomodate every kind of hardware. This library also has been wrapped and tested with OWB which offers internal API for spellchecking.
Download :
SpellChecker-1.0.lha (14KB)
And to see the library in real world usage:
http://fabportnawak.free.fr/temp/marked_errors.pnghttp://fabportnawak.free.fr/temp/suggestions.png
More info in the comments here:
http://www.morphzone.org/news/1857.html .