[libvoikko] Status of experimental backends
hatapitk at iki.fi
Sun May 2 19:51:48 EEST 2010
On Saturday 24 April 2010, Harri Pitkänen wrote:
> To help me make the right decision I'd like to ask everyone who has ever
> tested libvoikko with some language other than Finnish to answer these
> 1) Which languages you tested?
> 2) Which backends and morphologies you used?
> 3) If there are Hunspell dictionaries available for these languages how
> did libvoikko compare to Hunspell in terms of quality of spell checking,
> speed or memory use?
> I'll publish a summary of all answers I get in a week or so.
And the results are:
- Icelandic with Lttoolbox backend:
Tested by me. I don't know Icelandic but testing with some text I found from
the net suggests that there are lots of valid words that the speller does not
accept. Spelling suggestions have not been tailored for Icelandic. Lttoolbox
library has namespace issues. I did not do any performance tests.
I'd say that the overall quality is not good enough to be considered for a
stable release yet.
- Sámi with HFST backend:
Not really tested, but should be close. Linguistic quality is good, but there
are licensing issues. There are no known technical problems. Comparison with
Hunspell has not been done yet.
Depending on whether the licensing issues can be solved it seems that
Sámi/HFST combination has a change of becoming a supported language for
libvoikko 3.1. Please report any progress and test results related to this.
More information about the Libvoikko