All right, so it is wise to prepare an interface for ever developing
search procedures leaving implementation for future versions?
Alexey Parshin wrote:
> I don't think now it is a time for it. May be later? For now, I'd like
> just
> to implement a concept. If we decide to develop it any further than
> just a
> list of words, then we can always re-index any object later.
> For instance, for now it may be enough just to have a word in the
> list. But,
> we can also have ether a counter (how many times word appears in the
> context) - that is still a compact words list. Or, we can have all the
> (meaningful) words in the content in the same order, so we can compute
> distance between words and make search understand that distance.
>
> 2006/11/16, Ilya A. Volynets-Evenbakh <ilya@total-knowledge.com>:
>>
>> Are there any plans on weighting the keywords?
>>
>> Alexey Parshin wrote:
>> > Here are my conclusions:
>> >
>> > 1) It is author's duty to maintain these keywords especially if
object
>> > has multiple versions. As far as I understand, the small number of
>> > keywords isn't sufficient, and the large number is impossible to
>> > maintain.
>> > 2) If the keywords are not taken from the text or something
displayed
>> > along with the text - they don't associate with the object as
possible
>> > search conditions.
>> >
>> > Conclusions:
>> > 1) We don't need an artificial set of keywords that are not a
part of
>> > the object.
>> > 2) Keywords should be generated automatically from the object text.
>> > 3) If author needs something that associates an object or its part
>> > with keywords, the author should provide a short description of the
>> > object or its part in the object text.
>> > 4) Auto-generation should happen automatically after any object
>> > content modification. Keywords themselves are not humanly editable.
>> >
>> > If we implement keywords this way, it would be zero efforts to
>> > maintain these keywords, and no logical inconsistencies when a
set of
>> > keywords doesn't match the object content.
>> >
>> > 2006/11/15, sergey@total-knowledge.com
>> > <mailto:sergey@total-knowledge.com> <sergey@total-knowledge.com
>> > <mailto:sergey@total-knowledge.com>>:
>> >
>> >
>> > > 5. What if UU creates its own keyword list, asks the
author to
>> > edit it
>> > > and does not let it go, until the author adds at least one
word
>> not
>> > > listed?
>> > >
>> > > 6. Shouldn't the keyword list to be structured in some way?
>> > >
>> > > 7. How keyword lists are updated? Depending on events in each
>> > category
>> > > the UMO belongs to? This does bring in the idea of structured
>> > keyword
>> > > list.
>> >
>> > Regarding structuring.
>> >
>> > 1. Automaticaly generated keywords for the current UMO.
>> > 2. All "current UMO"'s children UMOs keywords(if any)
>> > 3. Author's custom keywords(if any)
>> >
>> > From UI point of view, all keywords will be displayed in the
same
>> text
>> > area Keywords.
>> > From DB point of view, it's up to Alexey to deside. Here is my
>> humble
>> > opinion:
>> > We may add keywords field to every *_content table where we may
>> store
>> > author's custom keywords, automaticaly generated keywords
will be
>> > stored
>> > in keywords_list table. This will allow keywords to be part of
the
>> > UMO
>> > versioning.
>> >
>> >
>> >
>> > > --
>> > >
>> > > Anatoly Volynets, Co-Founder
>> > > total-knowledge.com <http://total-knowledge.com>
>> > > culturedialogue.org <http://culturedialogue.org>
>> > >
>> > >
>> >
>> >
>> >
>> >
>> >
>> > --
>> > Alexey Parshin,
>> > http://www.sptk.net
>>
>> --
>> Ilya A. Volynets-Evenbakh
>> Total Knowledge. CTO
>> http://www.total-knowledge.com
>>
>>
>
>
--
Anatoly Volynets, Co-Founder
total-knowledge.com
culturedialogue.org