I wonder what's the best strategy in analyzing/indexing a text if I want to
have these search options (and any combinations of them) later on the index:
- Search for case-sensitive terms as well as case-insensitive terms
- Search for wildcard terms as well as exact terms
- Search for phrase terms
What I've read so far is that I usually need to create a new field with a
dedicated analyzer for every search option. E.g., if I want case-sensitive
and case-insensitive search then I have to index the text twice, one time
in a case-sensitive analyzed field and one time in a case-insensitive
analyzed field. Same approach goes for wildcards (realized through n-grams)
and exact matches (without n-grams). Is this correct so far?
Does this mean, that if I want any combinations of the mentioned search
options, then I have to create a new field for every combination? How do I
search for a combination then, like for "term1 (case-sensitive) AND term2
(case-insensitve)"?
--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
A multifield mapping would be best. One copy of the data can be indexed
several different ways. So for example, your query would specify:
myfield.case_insensitive:foo
myfield.case_sensitive:Foo
And so on. Of course, you can choose shorter names. And one of the mappings
can be the default, so for example you might wish that myfield is the
default assigned to case-insensitive queries.
(I tried to quickly look this up on the on-line guide, then fell back to my
own docs. That book will be very nice!!!)
On Tuesday, August 6, 2013 2:36:05 AM UTC-4, Stefan Pi wrote:
I wonder what's the best strategy in analyzing/indexing a text if I want
to have these search options (and any combinations of them) later on the
index:
- Search for case-sensitive terms as well as case-insensitive terms
- Search for wildcard terms as well as exact terms
- Search for phrase terms
What I've read so far is that I usually need to create a new field with a
dedicated analyzer for every search option. E.g., if I want case-sensitive
and case-insensitive search then I have to index the text twice, one time
in a case-sensitive analyzed field and one time in a case-insensitive
analyzed field. Same approach goes for wildcards (realized through n-grams)
and exact matches (without n-grams). Is this correct so far?
Does this mean, that if I want any combinations of the mentioned search
options, then I have to create a new field for every combination? How do I
search for a combination then, like for "term1 (case-sensitive) AND term2
(case-insensitve)"?
--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
But with this approach, how do I search for this phrase: "term1 term2*"?
the first term (term1) should match as it is (exact match) whereas the
second term should match any terms starting with term2 (wildcard match)
plus the second term must be the successor of the first in the original
text (phrase match)? I don't see how this can be arranged with multi
fields. Any ideas?
Thank you!
Am Dienstag, 6. August 2013 16:19:42 UTC+2 schrieb InquiringMind:
A multifield mapping would be best. One copy of the data can be indexed
several different ways. So for example, your query would specify:
myfield.case_insensitive:foo
myfield.case_sensitive:Foo
And so on. Of course, you can choose shorter names. And one of the
mappings can be the default, so for example you might wish that myfieldis the default assigned to case-insensitive queries.
(I tried to quickly look this up on the on-line guide, then fell back to
my own docs. That book will be very nice!!!)
On Tuesday, August 6, 2013 2:36:05 AM UTC-4, Stefan Pi wrote:
I wonder what's the best strategy in analyzing/indexing a text if I want
to have these search options (and any combinations of them) later on the
index:
- Search for case-sensitive terms as well as case-insensitive terms
- Search for wildcard terms as well as exact terms
- Search for phrase terms
What I've read so far is that I usually need to create a new field with a
dedicated analyzer for every search option. E.g., if I want case-sensitive
and case-insensitive search then I have to index the text twice, one time
in a case-sensitive analyzed field and one time in a case-insensitive
analyzed field. Same approach goes for wildcards (realized through n-grams)
and exact matches (without n-grams). Is this correct so far?
Does this mean, that if I want any combinations of the mentioned search
options, then I have to create a new field for every combination? How do I
search for a combination then, like for "term1 (case-sensitive) AND term2
(case-insensitve)"?
--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.
Stefan,
You are running into a Lucene limitation. It does not permit wildcards in
phrases.
In the long term, I hope to get the chance to teach Lucene how to do this
and more.
In the mean time, you could issue some form of AND or AND-like query (for
example, BoolQuery with must clauses for the individual terms). Then you
could post-process the response documents by implementing a recursive
algorithm that would be able to match phrases and wildcards. I've once
implemented this in C++ as a post-query filter, and it wouldn't be too
difficult to reincarnate it in Java.
But it would be faster if the Lucene index supported word position
natively. Then it could easily match phrases within the index, match
wildcard terms in phrases within the index, and use leapfrog logic to make
it all run blindingly fast. {"dream": "off" }
Brian
On Thursday, August 8, 2013 3:30:19 AM UTC-4, Stefan Pi wrote:
But with this approach, how do I search for this phrase: "term1 term2*"?
the first term (term1) should match as it is (exact match) whereas the
second term should match any terms starting with term2 (wildcard match)
plus the second term must be the successor of the first in the original
text (phrase match)? I don't see how this can be arranged with multi
fields. Any ideas?
Thank you!
--
You received this message because you are subscribed to the Google Groups "elasticsearch" group.
To unsubscribe from this group and stop receiving emails from it, send an email to elasticsearch+unsubscribe@googlegroups.com.
For more options, visit https://groups.google.com/groups/opt_out.