According to Geoff Hutchison:
> >5. I see the application of Fuzzy filters to lookups as an open issue.
> >Current fuzzy algos can be modeled as operators that result in a 'or' query
> >of single words. Is this optimum? Wouldn't it be perferable to have 'leaf'
> >queries (inherited from WordQuery) that do 'specialized' lookups according
> >to each fuzzy algo, then merged with a single 'or'? Dunno.
>
> My idea was that Fuzzy would be passed down the tree so that
> individual types (i.e. phrase/Exact) can ignore all or some fuzzies.
> When they hit the leafs, these would become OrParseTree objects and
> add the appropriate alternates. I'm open to suggestions on that.
>
> >6. (I'm not sure whether the following applies only to htsearch)
> >Altavista disposes of a powerful hypenated-word query similar to the phrase
> >query, so that:
> > foo-bar is equivalent to "foo bar" or "foo-bar" or "foobar"
> >but also:
>
> This could be a useful fuzzy algorithm. Right now they're not
> phrase-aware. my original idea is that Fuzzy objects can return a
> ParseTree of alternatives, so they can return phrases, normal
> OrParseTrees, etc. This would help for the synonym fuzzy too.
Since we're discussing the new ParseTree handling of fuzzy matches, I
thought I'd reiterate my idea about chaining or cascading fuzzy matches,
so that for example the endings algorithm could be applied to fuzzy
matches from synonyms, or vice-versa, or the accents algorithm could be
applied to matches from endings, to get a greater assortment of possible
fuzzy matches than the current approach. There would need to be a way
of controlling this, because in some cases you wouldn't want to chain
them, e.g. substring fuzzy matches shouldn't be further expanded by other
algorithms.
--
Gilles R. Detillieux E-mail: <[EMAIL PROTECTED]>
Spinal Cord Research Centre WWW: http://www.scrc.umanitoba.ca/~grdetil
Dept. Physiology, U. of Manitoba Phone: (204)789-3766
Winnipeg, MB R3E 3J7 (Canada) Fax: (204)789-3930
------------------------------------
To unsubscribe from the htdig3-dev mailing list, send a message to
[EMAIL PROTECTED]
You will receive a message to confirm this.
