Re: beagle r4293 - in trunk/beagle: BeagleClient beagled search search/Pages

[Debajyoti Bera <dbera web gmail com>]

> > Implement suggestions in an unobtrusive way using FuzzyTermEnum.
> > Suggestions are only generated upon request.
>
> How expensive is the search for suggestions?  If they're inexpensive,
> we might want to consider returning a SuggestionsResponse for every
> Query instead of or in addition to explicitly requesting them.

I have a general question on generating "suggestions"; I emailed Lukas but I 
will also ask here in case someone familiar with language processing knows 
the answer.

We only index the stemmed form of any word. We dont store the actual word 
anywhere. E.g. "dancing very nicely" gets indexed as "danc veri nice". Using 
the stemmed words to find similar words will only give us similar stems - 
which is OK but how to get an actual sensible English word from the stemmed 
word ? E.g. say the incorrect query was "vedy" - it will stem to "vedi" - 
thay will return a similar term "veri" - but how to get back "vedy" (or some 
meaningful word, need not be of the exact form as "very") ?

Are there known solutions without using wordlists ?

- dBera

-- 
-----------------------------------------------------
Debajyoti Bera @ http://dtecht.blogspot.com
beagle / KDE fan
Mandriva / Inspiron-1100 user