Re: Excalibur text datablade problems
Posted in 1998
In article <3614C27F.A25AF788@mrc.ac.za>,
Peter Gelderbloem <peter.gelderbloem@mrc.ac.za> wrote:
> i'm trying to do a text search on abstracts.i'm using ius with the
> excalibur textdatablade ( ETX.1.10.UC1 ).the index declaration is as
> follows:
>
> CREATE INDEX publication_idx2 ON publications (abstract
> etx_clob_ops)
> USING etx (WORD_SUPPORT = 'PATTERN', PHRASE_SUPPORT='MEDIUM');>
> and my query's where clause is:
>
> etx_contains(abstract,,Row('$keywords','PATTERN_BASIC'))
>
> as far as i know this is supposed to do super- and substring matches and
> not substitution or transposition matches.in practice it actually does
> this.
>
> anybody know how to fix this?
Today I'm using ETX.1.10.UC3 and to quote from the manual:
>>If you want to enable basic pattern matching, use the PATTERN_BASIC tuning
>>parameter. A search with this parameter will return the best matches that
>>have a score of WORD_SCORE or better. The search might return transposi-tions,
>>substitutions, and superstring and substring pattern matches
>>depending on the value of WORD_SCORE, though it is not guaranteed.
So, I'm not sure if you can just use superstring and substring pattern matches
since depending on WORD_SCORE it might use transpositions and
substitutions matching as well. Try setting WORD_SCORE (I think you can do
that in UC1) to something HIGHER than the default of 70. That might help you
out. I just tried it and with the data I have, a word score of 76 seemed to
turn of some of the transpositions. But your data is different so I'm betting
your # will be different. I'm not sure if there is a specific value to turn
it off. That's a good question though.
You could always switch to exact word matching. :-) But then you loose the
ability to look for substrings.
Tony
--
asalerno@monmouth.com
-----------== Posted via Deja News, The Discussion Network ==----------
http://www.dejanews.com/ Search, Read, Discuss, or Start Your Own