Follow-up to my problem searching the OPAC for phrases containing punctuation (i.e.,
Electroactive polymer (EAP) actuators). This Bywater
Solutions<https://bywatersolutions.com/education/elastic-searching> article
suggests that the problem is a feature of Elasticsearch.
FYI, we are using Elasticsearch 6.1.1 on its own dedicated server, and I don’t
believe we’ve installed the ICU Analysis plug-in (looks like it’s required for
Zebra, but I can’t tell if it’s required for Elasticsearch), which could be a
factor. I couldn’t replicate all aspects of my experience in a sandbox,
although searches for phrases containing punctuation still failed in an OPAC
“Library Catalog” sandbox search. I concluded that I needed to review our
configuration.
I’ve been reviewing the Koha Wiki and elastic.co documentation, and comparing
to our index_config.yaml.
By chance does anyone know how to interpret the syntax below? The
documentation describes the parameters below, but I don’t see any usage of “-“
before options. Does the option “- punctuation“ mean “yes, remove
punctuation”, or “no, don’t remove punctuation”, or does the phrase refer to
some additional configuration file, or perhaps it’s commented out?
analyzer_phrase:
tokenizer: keyword
filter:
- icu_folding
char_filter:
- punctuation
Thanks for your time and consideration,
Tasha Bales
Business Support Team | Information Services
Enterprise Services | Enterprise Operations, Finance and Sustainability
-----Original Message-----
From: Bales (US), Tasha R
Sent: Tuesday, June 22, 2021 7:54 AM
To: 'Jonathan Druart' <[email protected]>
Cc: Discussion Group Koha <[email protected]>
Subject: RE: [EXTERNAL] Re: [Koha] Title search works, but Library catalog
search fails in OPAC
Jonathan, thank you!
It does work without the parentheses.
I would suspect an encoding problem, but for that the problem only manifests in
the OPAC, and not the intranet.
I came across this issue while testing after migrating from MariaDB to Percona
MySQL. Your reply prompted me to check the encoding of the new database, and
it's unfortunately Latin-1. Since these are parentheses and not diacritics,
I’m not sure what my expectations should be, but changing to UTF-8 is a place
to start. Httpd.conf does have UTF-8 set as the default.
FWIW, my source records were encoded in MARC-8. I used MarcEdit to convert
them to UTF-8, and it appears that Koha automatically converts anyway on
import. When I loaded these records into Koha, I used bulkmarcimport.pl on the
command line.
I'll ask that the default character set of the database be changed, and see if
that helps. Thanks again. I'm embarrassed that I didn't think to omit the
parentheses, or rather was belligerently insisting to myself that they should
not have been a problem,
Tasha Bales
Business Support Team | Information Services Enterprise Services | Enterprise
Operations, Finance and Sustainability
(480) 509-5415
https://is.web.boeing.com
-----Original Message-----
From: Jonathan Druart [mailto:[email protected]]
Sent: Tuesday, June 22, 2021 12:01 AM
To: Bales (US), Tasha R <[email protected]>
Cc: Discussion Group Koha <[email protected]>
Subject: [EXTERNAL] Re: [Koha] Title search works, but Library catalog search
fails in OPAC
Importance: High
EXT email: be mindful of links/attachments.
Hello Tasha,
I've created 2 records with
245$a Electroactive polymer (EAP) actuators as artificial muscles and the
following query returns the 2 results.
/opac-search.pl?idx=&q=Electroactive%20polymer%20%28EAP%29%20actuators&weight_search=1
Tried on master and 20.11.06.
Maybe a silly idea: does it work without the parenthesis?
Could you try and recreate it on a sandbox
(https://wiki.koha-community.org/wiki/Sandboxes) and provide us a step by step
plan to reproduce the problem?
Regards,
Jonathan
Le mar. 22 juin 2021 à 00:46, Bales (US), Tasha R <[email protected]> a
écrit :
Good afternoon,
I’m having trouble with Title vs. Library catalog keyword searching with
several example titles. Searching the same phrase with either method yields
different results. This problem occurs only in the OPAC. I hope to confirm
whether the behavior I’m seeing is intended (i.e., the problem is me) or not.
Thanks in advance.
For example, given the ebook title, Electroactive polymer (EAP)
actuators as artificial muscles, a Title keyword search in the OPAC is
successful, but a plain, Library catalog (i.e., no index specified), keyword
search fails.
For reference, the title is recorded in the MARC record as:
245 00 - TITLE STATEMENT
a Title Electroactive polymer (EAP) actuators as artificial muscles :
Below I’ve copied in my search history as well as the tail of the search URL
that shows the search parameters.
· Library catalog keyword search with 0 results
o 2021-06-21 02:34 PM Electroactive polymer (EAP) actuators,
suppress:false 0
o
…opac-search.pl?idx=&q=Electroactive%20polymer%20%28EAP%29%20actuators&weight_search=1
· Title keyword search with 2 results
o 2021-06-21 02:34 PM Electroactive polymer (EAP) actuators,
suppress:false 2
o
…opac-search.pl?idx=ti&q=Electroactive+polymer+%28EAP%29+actuators&weight_search=1
As a test, I decided to enclose my Library catalog search terms in quotes,
which yielded the desired results. However, I did not at all anticipate that
quotes would be required to get hits:
· Library catalog quoted keyword search with 2 results
o 2021-06-21 02:46 PM "Electroactive polymer (EAP) actuators",
suppress:false 2
o …
opac-search.pl?idx=&q=%22Electroactive+polymer+%28EAP%29+actuators%22&weight_search=1
On comparing the above URL query strings, it appears that the unquoted terms in
the Library catalog keyword search aren’t “anded” together with a “+” the way
other searches are, but I’m not sure what the implications are, if any. Also,
the Koha manual indicates the following, which suggests to me that I ought to
get hits on the unquoted string:
When you have more than one word in the search box, Koha will still do a
keyword search, but a bit differently. Each word will be searched on its own,
then the Boolean connector ‘and’ will narrow your search to those items with
all words contained in matching records.
I understand and can predict pretty well the way our old ILS (Millennium, if
context helps) will perform a keyword search, but I’m a little confused here.
My expectation for this particular case is that all of the above methods would
yield results. If there are any pointers to be had, I thank you if might point
me to them so that I may be better poised to help users.
I’m using Elasticsearch with Koha 20.11.06. I reindexed both authorities and
biblios today, but that didn’t impact my experience. The records are not newly
added.
Thanks!
Tasha Bales
Business Support Team | Information Services Enterprise Services |
Enterprise Operations, Finance and Sustainability
_______________________________________________
Koha mailing list http://koha-community.org [email protected]
Unsubscribe: https://lists.katipo.co.nz/mailman/listinfo/koha
_______________________________________________
Koha mailing list http://koha-community.org
[email protected]
Unsubscribe: https://lists.katipo.co.nz/mailman/listinfo/koha
